Showing 1–50 of 84 results for author: Hosseini, A

TokenSwap: Benchmarking and Reducing the Modality Gap in Multimodal LLMs

Authors: Andong Hua, Colton Bishop, Igor Mordatch, Arian Hosseini, Jindong Gu, Aleksandra Faust, Rebecca Roelofs, Yao Qin

Abstract: Multimodal large language models (MLLMs) should generate consistent responses given semantically equivalent inputs across modalities. However, we observe a systematic discrepancy in model predictions under such cross-modal variations. Specifically, we define the modality gap as the difference in model performance under semantically equivalent textual and multimodal inputs. We introduce TokenSwap,… ▽ More Multimodal large language models (MLLMs) should generate consistent responses given semantically equivalent inputs across modalities. However, we observe a systematic discrepancy in model predictions under such cross-modal variations. Specifically, we define the modality gap as the difference in model performance under semantically equivalent textual and multimodal inputs. We introduce TokenSwap, a method that constructs such inputs by replacing textual concepts with semantically aligned images, resulting in sequences where visual tokens are interleaved with text tokens. Based on TokenSwap, we transform existing text-based benchmarks such as MMLU into image-interleaved counterparts, resulting in TokenSwap-Bench. Across 42 MLLMs, we observe a pervasive modality gap, with performance decreasing by 4.2% to 47.4% when moving from text-only to image-interleaved inputs, averaging 19.6% +/- 3.3% across models. Notably, we observe that reasoning models exhibit consistently smaller gaps, achieving an average gap of 10.1% compared to 25.5% for non-reasoning models. In contrast, neither prompting strategies nor scaling training compute alone reliably reduces the modality gap. Finally, we demonstrate that incorporating TokenSwap during training effectively mitigates this gap while preserving strong text-only and vision-language performance. △ Less

Submitted 19 May, 2026; originally announced July 2026.

Gemma 4 Technical Report

Authors: Gemma Team, Sherif El Abd, Vaibhav Aggarwal, Robin Algayres, Alek Andreev, Olivier Bachem, Ian Ballantyne, Cormac Brick, Victor Cărbune, Michelle Casbon, Mayank Chaturvedi, Aditya Chawla, Victor Cotruta, Alice Coucke, Phil Culliton, Robert Dadashi, Lucas Dixon, Mohamed Elhawaty, Utku Evci, Clément Farabet, Johan Ferret, Filippo Galgani, Sertan Girgin, Jean-Bastien Grill, Maarten Grootendorst , et al. (298 additional authors not shown)

Abstract: We introduce Gemma 4, a new generation of open-weight, natively multimodal language models in the Gemma model family. Designed to advance compute efficiency and reasoning, the Gemma 4 model suite features dense and Mixture-of-Experts architectures, ranging from 2.3B to 31B parameters. Alongside improved vision and audio encoders for all model sizes, we propose a unified, encoder-free architecture… ▽ More We introduce Gemma 4, a new generation of open-weight, natively multimodal language models in the Gemma model family. Designed to advance compute efficiency and reasoning, the Gemma 4 model suite features dense and Mixture-of-Experts architectures, ranging from 2.3B to 31B parameters. Alongside improved vision and audio encoders for all model sizes, we propose a unified, encoder-free architecture for our 12B model, which ingests raw audio and image patches. Furthermore, we integrate a thinking mode, enabling Gemma models to generate reasoning traces prior to responding. We improve inference speed, memory, and compute efficiency, as well as long-context abilities through critical design choices. Gemma 4 establishes a leap in performance across STEM, multimodal, and long-context benchmarks, and rivals larger, frontier open models in human-rated tasks. △ Less

Submitted 24 July, 2026; v1 submitted 2 July, 2026; originally announced July 2026.

Comments: 17 pages, 2 figures, technical report, updated

Boundary condition fidelity for bottom-hole pressure and CO2 plume prediction in geological carbon storage

Authors: Romal Ramadhan, Seyyed A. Hosseini, Larry W. Lake

Abstract: Accurate prediction of bottom-hole pressure (BHP) and CO2 plume migration is essential for safe geological carbon storage, yet practical simulations often rely on truncated domains where artificial boundaries distort pressure diffusion and CO2 saturation footprints. In this study, we evaluate how boundary-condition fidelity affects BHP and CO2 plume prediction by comparing ten reduced-domain bound… ▽ More Accurate prediction of bottom-hole pressure (BHP) and CO2 plume migration is essential for safe geological carbon storage, yet practical simulations often rely on truncated domains where artificial boundaries distort pressure diffusion and CO2 saturation footprints. In this study, we evaluate how boundary-condition fidelity affects BHP and CO2 plume prediction by comparing ten reduced-domain boundary treatments against full-domain reference simulations in homogeneous and heterogeneous reservoirs. We test uniform pore-volume multipliers, transmissibility modifiers, corner-adjusted pore-volume corrections, layered corrections, and gradual modifiers using BHP RMSE, NRMSE, peak pressure deviation, and plume Intersection over Union (IoU) as performance metrics. Our results show that conserving corner pore volume is the most important requirement for truncated-domain modeling. We find that uniform treatments which neglect corner storage generate large pressure errors, with BHP RMSE of 362 to 382 psi in the homogeneous model and 250 to 304 psi in the heterogeneous model, and yield plume IoU values near 0.80 to 0.84, indicating roughly 16 to 20% of the combined plume area is misrepresented. Corner-adjusted scenarios substantially reduce pressure errors and raise plume IoU above 0.94, but we observe that transmissibility correction is not universally beneficial. In homogeneous reservoirs, uniform transmissibility adjustment improves pressure fidelity; in heterogeneous reservoirs, it can over-restrict flow across variable-permeability boundary faces, increasing BHP error and contracting the predicted plume. We find the gradual modifier with transmissibility correction provides the most consistent performance, achieving BHP NRMSE below 3.7% and plume IoU above 0.97 in both reservoir types. △ Less

Submitted 25 June, 2026; originally announced June 2026.

Representational Curvature Modulates Behavioral Uncertainty in Large Language Models

Authors: Jack King, Evelina Fedorenko, Eghbal A. Hosseini

Abstract: In autoregressive large language models (LLMs), temporal straightening offers an account of how the next-token prediction objective shapes representations. Models learn to progressively straighten the representational trajectory of input sequences across layers, potentially facilitating next-token prediction via linear extrapolation. However, a direct link between this trajectory and token-level b… ▽ More In autoregressive large language models (LLMs), temporal straightening offers an account of how the next-token prediction objective shapes representations. Models learn to progressively straighten the representational trajectory of input sequences across layers, potentially facilitating next-token prediction via linear extrapolation. However, a direct link between this trajectory and token-level behavior has been missing. We provide such a link by relating contextual curvature-a geometric measure of how sharply the representational trajectory bends over recent context-to next-token entropy. Across two models (GPT-2 XL and Pythia-2.8B), contextual curvature is correlated with entropy, and this relationship emerges during training. Perturbation experiments reveal selective dependence: manipulating curvature through trajectory-aligned interventions reliably modulates entropy, while geometrically misaligned perturbations have no effect. Finally, regularizing representations to be straighter during training modestly reduces token-level entropy without degrading validation loss. These results identify trajectory curvature as a task-aligned representational feature that influences behavioral uncertainty in LLMs. △ Less

Submitted 26 April, 2026; originally announced April 2026.

ReLIC-SGG: Relation Lattice Completion for Open-Vocabulary Scene Graph Generation

Authors: Amir Hosseini, Sara Farahani, Xinyi Li, Suiyang Guang

Abstract: Open-vocabulary scene graph generation (SGG) aims to describe visual scenes with flexible relation phrases beyond a fixed predicate set. Existing methods usually treat annotated triplets as positives and all unannotated object-pair relations as negatives. However, scene graph annotations are inherently incomplete: many valid relations are missing, and the same interaction can be described at diffe… ▽ More Open-vocabulary scene graph generation (SGG) aims to describe visual scenes with flexible relation phrases beyond a fixed predicate set. Existing methods usually treat annotated triplets as positives and all unannotated object-pair relations as negatives. However, scene graph annotations are inherently incomplete: many valid relations are missing, and the same interaction can be described at different granularities, e.g., \textit{on}, \textit{standing on}, \textit{resting on}, and \textit{supported by}. This issue becomes more severe in open-vocabulary SGG due to the much larger relation space. We propose \textbf{ReLIC-SGG}, a relation-incompleteness-aware framework that treats unannotated relations as latent variables rather than definite negatives. ReLIC-SGG builds a semantic relation lattice to model similarity, entailment, and contradiction among open-vocabulary predicates, and uses it to infer missing positive relations from visual-language compatibility, graph context, and semantic consistency. A positive-unlabeled graph learning objective further reduces false-negative supervision, while lattice-guided decoding produces compact and semantically consistent scene graphs. Experiments on conventional, open-vocabulary, and panoptic SGG benchmarks show that ReLIC-SGG improves rare and unseen predicate recognition and better recovers missing relations. △ Less

Submitted 25 May, 2026; v1 submitted 24 April, 2026; originally announced April 2026.

Comments: Some errors in the experimental sections

Modulating Cross-Modal Convergence with Single-Stimulus, Intra-Modal Dispersion

Authors: Eghbal A. Hosseini, Brian Cheung, Evelina Fedorenko, Alex H. Williams

Abstract: Neural networks exhibit a remarkable degree of representational convergence across diverse architectures, training objectives, and even data modalities. This convergence is predictive of alignment with brain representation. A recent hypothesis suggests this arises from learning the underlying structure in the environment in similar ways. However, it is unclear how individual stimuli elicit converg… ▽ More Neural networks exhibit a remarkable degree of representational convergence across diverse architectures, training objectives, and even data modalities. This convergence is predictive of alignment with brain representation. A recent hypothesis suggests this arises from learning the underlying structure in the environment in similar ways. However, it is unclear how individual stimuli elicit convergent representations across networks. An image can be perceived in multiple ways and expressed differently using words. Here, we introduce a methodology based on the Generalized Procrustes Algorithm to measure intra-modal representational convergence at the single-stimulus level. We applied this to vision models with distinct training objectives, selecting stimuli based on their degree of alignment (intra-modal dispersion). Crucially, we found that this intra-modal dispersion strongly modulates alignment between vision and language models (cross-modal convergence). Specifically, stimuli with low intra-modal dispersion (high agreement among vision models) elicited significantly higher cross-modal alignment than those with high dispersion, by up to a factor of two (e.g., in pairings of DINOv2 with language models). This effect was robust to stimulus selection criteria and generalized across different pairings of vision and language models. Measuring convergence at the single-stimulus level provides a path toward understanding the sources of convergence and divergence across modalities, and between neural networks and human neural representations. △ Less

Submitted 23 April, 2026; originally announced April 2026.

Journal ref: ICLR 2026 Workshop on Representational Alignment (Re-Align)

Context Structure Reshapes the Representational Geometry of Language Models

Authors: Eghbal A. Hosseini, Yuxuan Li, Yasaman Bahri, Declan Campbell, Andrew Kyle Lampinen

Abstract: Large Language Models (LLMs) have been shown to organize the representations of input sequences into straighter neural trajectories in their deep layers, which has been hypothesized to facilitate next-token prediction via linear extrapolation. Language models can also adapt to diverse tasks and learn new structure in context, and recent work has shown that this in-context learning (ICL) can be ref… ▽ More Large Language Models (LLMs) have been shown to organize the representations of input sequences into straighter neural trajectories in their deep layers, which has been hypothesized to facilitate next-token prediction via linear extrapolation. Language models can also adapt to diverse tasks and learn new structure in context, and recent work has shown that this in-context learning (ICL) can be reflected in representational changes. Here we bring these two lines of research together to explore whether representation straightening occurs \emph{within} a context during ICL. We measure representational straightening in Gemma 2 models across a diverse set of in-context tasks, and uncover a dichotomy in how LLMs' representations change in context. In continual prediction settings (e.g., natural language, grid world traversal tasks) we observe that increasing context increases the straightness of neural sequence trajectories, which is correlated with improvement in model prediction. Conversely, in structured prediction settings (e.g., few-shot tasks), straightening is inconsistent -- it is only present in phases of the task with explicit structure (e.g., repeating a template), but vanishes elsewhere. These results suggest that ICL is not a monolithic process. Instead, we propose that LLMs function like a Swiss Army knife: depending on task structure, the LLM dynamically selects between strategies, only some of which yield representational straightening. △ Less

Submitted 29 January, 2026; originally announced January 2026.

Trust, Don't Trust, or Flip: Robust Preference-Based Reinforcement Learning with Multi-Expert Feedback

Authors: Seyed Amir Hosseini, Maryam Abdolali, Amirhosein Tavakkoli, Fardin Ayar, Ehsan Javanmardi, Manabu Tsukada, Mahdi Javanmardi

Abstract: Preference-based reinforcement learning (PBRL) offers a promising alternative to explicit reward engineering by learning from pairwise trajectory comparisons. However, real-world preference data often comes from heterogeneous annotators with varying reliability; some accurate, some noisy, and some systematically adversarial. Existing PBRL methods either treat all feedback equally or attempt to fil… ▽ More Preference-based reinforcement learning (PBRL) offers a promising alternative to explicit reward engineering by learning from pairwise trajectory comparisons. However, real-world preference data often comes from heterogeneous annotators with varying reliability; some accurate, some noisy, and some systematically adversarial. Existing PBRL methods either treat all feedback equally or attempt to filter out unreliable sources, but both approaches fail when faced with adversarial annotators who systematically provide incorrect preferences. We introduce TriTrust-PBRL (TTP), a unified framework that jointly learns a shared reward model and expert-specific trust parameters from multi-expert preference feedback. The key insight is that trust parameters naturally evolve during gradient-based optimization to be positive (trust), near zero (ignore), or negative (flip), enabling the model to automatically invert adversarial preferences and recover useful signal rather than merely discarding corrupted feedback. We provide theoretical analysis establishing identifiability guarantees and detailed gradient analysis that explains how expert separation emerges naturally during training without explicit supervision. Empirically, we evaluate TTP on four diverse domains spanning manipulation tasks (MetaWorld) and locomotion (DM Control) under various corruption scenarios. TTP achieves state-of-the-art robustness, maintaining near-oracle performance under adversarial corruption while standard PBRL methods fail catastrophically. Notably, TTP outperforms existing baselines by successfully learning from mixed expert pools containing both reliable and adversarial annotators, all while requiring no expert features beyond identification indices and integrating seamlessly with existing PBRL pipelines. △ Less

Submitted 26 January, 2026; originally announced January 2026.

Comments: Equal contribution: Seyed Amir Hosseini and Maryam Abdolali. Corresponding author: Maryam Abdolali (maryam.abdolali@kntu.ac.ir)

Shape of Thought: When Distribution Matters More than Correctness in Reasoning Tasks

Authors: Abhranil Chandra, Ayush Agrawal, Arian Hosseini, Sebastian Fischmeister, Rishabh Agarwal, Navin Goyal, Aaron Courville

Abstract: We present the surprising finding that a language model's reasoning capabilities can be improved by training on synthetic datasets of chain-of-thought (CoT) traces from more capable models, even when all of those traces lead to an incorrect final answer. Our experiments show this approach can yield better performance on reasoning tasks than training on human-annotated datasets. We hypothesize that… ▽ More We present the surprising finding that a language model's reasoning capabilities can be improved by training on synthetic datasets of chain-of-thought (CoT) traces from more capable models, even when all of those traces lead to an incorrect final answer. Our experiments show this approach can yield better performance on reasoning tasks than training on human-annotated datasets. We hypothesize that two key factors explain this phenomenon: first, the distribution of synthetic data is inherently closer to the language model's own distribution, making it more amenable to learning. Second, these `incorrect' traces are often only partially flawed and contain valid reasoning steps from which the model can learn. To further test the first hypothesis, we use a language model to paraphrase human-annotated traces -- shifting their distribution closer to the model's own distribution -- and show that this improves performance. For the second hypothesis, we introduce increasingly flawed CoT traces and study to what extent models are tolerant to these flaws. We demonstrate our findings across various reasoning domains like math, algorithmic reasoning and code generation using MATH, GSM8K, Countdown and MBPP datasets on various language models ranging from 1.5B to 9B across Qwen, Llama, and Gemma models. Our study shows that curating datasets that are closer to the model's distribution is a critical aspect to consider. We also show that a correct final answer is not always a reliable indicator of a faithful reasoning process. △ Less

Submitted 22 January, 2026; v1 submitted 24 December, 2025; originally announced December 2025.

Real-Time Semantic Segmentation on FPGA for Autonomous Vehicles Using LMIINet with the CGRA4ML Framework

Authors: Amir Mohammad Khadem Hosseini, Sattar Mirzakuchaki

Abstract: Semantic segmentation has emerged as a fundamental problem in computer vision, gaining particular importance in real-time applications such as autonomous driving. The main challenge is achieving high accuracy while operating under computational and hardware constraints. In this research, we present an FPGA-based implementation of real-time semantic segmentation leveraging the lightweight LMIINet a… ▽ More Semantic segmentation has emerged as a fundamental problem in computer vision, gaining particular importance in real-time applications such as autonomous driving. The main challenge is achieving high accuracy while operating under computational and hardware constraints. In this research, we present an FPGA-based implementation of real-time semantic segmentation leveraging the lightweight LMIINet architecture and the Coarse-Grained Reconfigurable Array for Machine Learning (CGRA4ML) hardware framework. The model was trained using Quantization-Aware Training (QAT) with 8-bit precision on the Cityscapes dataset, reducing memory footprint by a factor of four while enabling efficient fixed-point computations. Necessary modifications were applied to adapt the model to CGRA4ML constraints, including simplifying skip connections, employing hardware-friendly operations such as depthwise-separable and 1A-1 convolutions, and redesigning parts of the Flatten Transformer. Our implementation achieves approximately 90% pixel accuracy and 45% mean Intersection-over-Union (mIoU), operating in real-time at 20 frames per second (FPS) with 50.1 ms latency on the ZCU104 FPGA board. The results demonstrate the potential of CGRA4ML, with its flexibility in mapping modern layers and off-chip memory utilization for skip connections, provides a path for implementing advanced semantic segmentation networks on FPGA for real-time applications to outperform traditional GPU solutions in terms of power efficiency while maintaining competitive accuracy. The code for this project is publicly available at https://github.com/STAmirr/ cgra4ml_semantic_segmentation △ Less

Submitted 25 October, 2025; originally announced October 2025.

Extending Resource Constrained Project Scheduling to Mega-Projects with Model-Based Systems Engineering & Hetero-functional Graph Theory

Authors: Amirreza Hosseini, Amro M. Farid

Abstract: Within the project management context, project scheduling serves as an indispensable component, functioning as a fundamental tool for planning, monitoring, controlling, and managing projects more broadly. Although the resource-constrained project scheduling problem (RCPSP) lies at the core of project management activities, it remains largely disconnected from the broader literature on model-based… ▽ More Within the project management context, project scheduling serves as an indispensable component, functioning as a fundamental tool for planning, monitoring, controlling, and managing projects more broadly. Although the resource-constrained project scheduling problem (RCPSP) lies at the core of project management activities, it remains largely disconnected from the broader literature on model-based systems engineering (MBSE), thereby limiting its integration into the design and management of complex systems. The original contribution of this paper is twofold. First, the paper seeks to reconcile the RCPSP with the broader literature and vocabulary of model-based systems engineering and hetero-functional graph theory (HFGT). A concrete translation pipeline from an activity-on-node network to a SysML activity diagram, and then to an operand net is constructed. Using this representation, it specializes the hetero-functional network minimum-cost flow (HFNMCF) formulation to the RCPSP context as a systematic means of HFGT for quantitative analysis and proves that the RCPSP is recoverable as a special case of a broader model. Secondly, on an illustrative instance with renewable and non-renewable operands, the specialized HFNMCF, while producing similar schedules, yields explicit explanations of the project states that enable richer monitoring and control. Overall, the framework preserves the strengths of the classical RCPSP while accommodating real-world constraints and enterprise-level decision processes encountered in large, complex megaprojects. △ Less

Submitted 21 October, 2025; originally announced October 2025.

Argos: A Decentralized Federated System for Detection of Traffic Signs in CAVs

Authors: Seyed Mahdi Haji Seyed Hossein, Alireza Hosseini, Soheil Hajian Manesh, Amirali Shahriary

Abstract: Connected and automated vehicles generate vast amounts of sensor data daily, raising significant privacy and communication challenges for centralized machine learning approaches in perception tasks. This study presents a decentralized, federated learning framework tailored for traffic sign detection in vehicular networks to enable collaborative model training without sharing raw data. The framewor… ▽ More Connected and automated vehicles generate vast amounts of sensor data daily, raising significant privacy and communication challenges for centralized machine learning approaches in perception tasks. This study presents a decentralized, federated learning framework tailored for traffic sign detection in vehicular networks to enable collaborative model training without sharing raw data. The framework partitioned traffic sign classes across vehicles for specialized local training using lightweight object detectors, aggregated model parameters via algorithms like FedProx, FedAdam and FedAVG in a simulated environment with the Flower framework, and evaluated multiple configurations including varying server rounds, local epochs, client participation fractions, and data distributions. Experiments demonstrated that increasing server rounds from 2 to 20 boosted accuracy from below 0.1 to over 0.8, moderate local epochs (8-10) provided optimal efficiency with accuracies around 0.67, higher client participation fractions enhanced generalization up to 0.83, FedProx outperformed other aggregators in handling heterogeneity, non-IID data distributions reduced performance compared to IID, and training duration primarily scaled with the number of rounds rather than aggregation strategy. We conclude that this federated approach may offer a scalable, privacy-preserving solution for real-world vehicular deployments, potentially guiding future integrations of robust aggregation and communication optimizations to advance intelligent transportation systems. △ Less

Submitted 18 August, 2025; originally announced August 2025.

Comments: 7 pages, 10 figures

ACM Class: I.2.6; I.4.8

Optimal CO2 storage management considering safety constraints in multi-stakeholder multi-site CCS projects: a Markov game perspective

Authors: Jungang Chen, Seyyed A. Hosseini

Abstract: Carbon capture and storage (CCS) projects typically involve a diverse array of stakeholders or players from public, private, and regulatory sectors, each with different objectives and responsibilities. Given the complexity, scale, and long-term nature of CCS operations, determining whether individual stakeholders can independently maximize their interests or whether collaborative coalition agreeme… ▽ More Carbon capture and storage (CCS) projects typically involve a diverse array of stakeholders or players from public, private, and regulatory sectors, each with different objectives and responsibilities. Given the complexity, scale, and long-term nature of CCS operations, determining whether individual stakeholders can independently maximize their interests or whether collaborative coalition agreements are needed remains a central question for effective CCS project planning and management. CCS projects are often implemented in geologically connected sites, where shared geological features such as pressure space and reservoir pore capacity can lead to competitive behavior among stakeholders. Furthermore, CO2 storage sites are often located in geologically mature basins that previously served as sites for hydrocarbon extraction or wastewater disposal in order to leverage existing infrastructures, which makes unilateral optimization even more complicated and unrealistic. In this work, we propose a paradigm based on Markov games to quantitatively investigate how different coalition structures affect the goals of stakeholders. We frame this multi-stakeholder multi-site problem as a multi-agent reinforcement learning problem with safety constraints. Our approach enables agents to learn optimal strategies while compliant with safety regulations. We present an example where multiple operators are injecting CO2 into their respective project areas in a geologically connected basin. To address the high computational cost of repeated simulations of high-fidelity models, a previously developed surrogate model based on the Embed-to-Control (E2C) framework is employed. Our results demonstrate the effectiveness of the proposed framework in addressing optimal management of CO2 storage when multiple stakeholders with various objectives and goals are involved. △ Less

Submitted 22 January, 2026; v1 submitted 15 August, 2025; originally announced August 2025.

Comments: 58 pages

Journal ref: Int. J. Greenh. Gas Control 149 (2026) 104683

Multi-Turn Puzzles: Evaluating Interactive Reasoning and Strategic Dialogue in LLMs

Authors: Kartikeya Badola, Jonathan Simon, Arian Hosseini, Sara Marie Mc Carthy, Tsendsuren Munkhdalai, Abhimanyu Goyal, Tomáš Kočiský, Shyam Upadhyay, Bahare Fatemi, Mehran Kazemi

Abstract: Large language models (LLMs) excel at solving problems with clear and complete statements, but often struggle with nuanced environments or interactive tasks which are common in most real-world scenarios. This highlights the critical need for developing LLMs that can effectively engage in logically consistent multi-turn dialogue, seek information and reason with incomplete data. To this end, we int… ▽ More Large language models (LLMs) excel at solving problems with clear and complete statements, but often struggle with nuanced environments or interactive tasks which are common in most real-world scenarios. This highlights the critical need for developing LLMs that can effectively engage in logically consistent multi-turn dialogue, seek information and reason with incomplete data. To this end, we introduce a novel benchmark comprising a suite of multi-turn tasks each designed to test specific reasoning, interactive dialogue, and information-seeking abilities. These tasks have deterministic scoring mechanisms, thus eliminating the need for human intervention. Evaluating frontier models on our benchmark reveals significant headroom. Our analysis shows that most errors emerge from poor instruction following, reasoning failures, and poor planning. This benchmark provides valuable insights into the strengths and weaknesses of current LLMs in handling complex, interactive scenarios and offers a robust platform for future research aimed at improving these critical capabilities. △ Less

Submitted 24 August, 2025; v1 submitted 13 August, 2025; originally announced August 2025.

Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities

Authors: Gheorghe Comanici, Eric Bieber, Mike Schaekermann, Ice Pasupat, Noveen Sachdeva, Inderjit Dhillon, Marcel Blistein, Ori Ram, Dan Zhang, Evan Rosen, Luke Marris, Sam Petulla, Colin Gaffney, Asaf Aharoni, Nathan Lintz, Tiago Cardal Pais, Henrik Jacobsson, Idan Szpektor, Nan-Jiang Jiang, Krishna Haridasan, Ahmed Omran, Nikunj Saunshi, Dara Bahri, Gaurav Mishra, Eric Chu , et al. (3410 additional authors not shown)

Abstract: In this report, we introduce the Gemini 2.X model family: Gemini 2.5 Pro and Gemini 2.5 Flash, as well as our earlier Gemini 2.0 Flash and Flash-Lite models. Gemini 2.5 Pro is our most capable model yet, achieving SoTA performance on frontier coding and reasoning benchmarks. In addition to its incredible coding and reasoning skills, Gemini 2.5 Pro is a thinking model that excels at multimodal unde… ▽ More In this report, we introduce the Gemini 2.X model family: Gemini 2.5 Pro and Gemini 2.5 Flash, as well as our earlier Gemini 2.0 Flash and Flash-Lite models. Gemini 2.5 Pro is our most capable model yet, achieving SoTA performance on frontier coding and reasoning benchmarks. In addition to its incredible coding and reasoning skills, Gemini 2.5 Pro is a thinking model that excels at multimodal understanding and it is now able to process up to 3 hours of video content. Its unique combination of long context, multimodal and reasoning capabilities can be combined to unlock new agentic workflows. Gemini 2.5 Flash provides excellent reasoning abilities at a fraction of the compute and latency requirements and Gemini 2.0 Flash and Flash-Lite provide high performance at low latency and cost. Taken together, the Gemini 2.X model generation spans the full Pareto frontier of model capability vs cost, allowing users to explore the boundaries of what is possible with complex agentic problem solving. △ Less

Submitted 19 December, 2025; v1 submitted 7 July, 2025; originally announced July 2025.

Comments: 72 pages, 17 figures

Efficient Malicious UAV Detection Using Autoencoder-TSMamba Integration

Authors: Azim Akhtarshenas, Ramin Toosi, David López-Pérez, Tohid Alizadeh, Alireza Hosseini

Abstract: Malicious Unmanned Aerial Vehicles (UAVs) present a significant threat to next-generation networks (NGNs), posing risks such as unauthorized surveillance, data theft, and the delivery of hazardous materials. This paper proposes an integrated (AE)-classifier system to detect malicious UAVs. The proposed AE, based on a 4-layer Tri-orientated Spatial Mamba (TSMamba) architecture, effectively captures… ▽ More Malicious Unmanned Aerial Vehicles (UAVs) present a significant threat to next-generation networks (NGNs), posing risks such as unauthorized surveillance, data theft, and the delivery of hazardous materials. This paper proposes an integrated (AE)-classifier system to detect malicious UAVs. The proposed AE, based on a 4-layer Tri-orientated Spatial Mamba (TSMamba) architecture, effectively captures complex spatial relationships crucial for identifying malicious UAV activities. The first phase involves generating residual values through the AE, which are subsequently processed by a ResNet-based classifier. This classifier leverages the residual values to achieve lower complexity and higher accuracy. Our experiments demonstrate significant improvements in both binary and multi-class classification scenarios, achieving up to 99.8 % recall compared to 96.7 % in the benchmark. Additionally, our method reduces computational complexity, making it more suitable for large-scale deployment. These results highlight the robustness and scalability of our approach, offering an effective solution for malicious UAV detection in NGN environments. △ Less

Submitted 14 May, 2025; originally announced May 2025.

Comments: 12 pages, 6 figures and 3 tables, accepted in IbPRIA 2025, https://www.ibpria.org/2025/?page=dates

Putting the Value Back in RL: Better Test-Time Scaling by Unifying LLM Reasoners With Verifiers

Authors: Kusha Sareen, Morgane M Moss, Alessandro Sordoni, Rishabh Agarwal, Arian Hosseini

Abstract: Prevalent reinforcement learning~(RL) methods for fine-tuning LLM reasoners, such as GRPO or Leave-one-out PPO, abandon the learned value function in favor of empirically estimated returns. This hinders test-time compute scaling that relies on using the value-function for verification. Yet if parallel test-time compute is already part of the deployment plan, training should be designed to support… ▽ More Prevalent reinforcement learning~(RL) methods for fine-tuning LLM reasoners, such as GRPO or Leave-one-out PPO, abandon the learned value function in favor of empirically estimated returns. This hinders test-time compute scaling that relies on using the value-function for verification. Yet if parallel test-time compute is already part of the deployment plan, training should be designed to support it. In this work, we propose RL$^V$ that augments any ``value-free'' RL method by jointly training the LLM as both a reasoner and a generative verifier using RL-generated data, adding verification capabilities without significant overhead. Empirically, RL$^V$ boosts MATH accuracy by over 20\% with parallel sampling and enables $8-32\times$ efficient test-time compute scaling compared to the base RL method. RL$^V$ also exhibits strong generalization capabilities for both easy-to-hard and out-of-domain tasks. Furthermore, RL$^V$ achieves $1.2-1.6\times$ higher performance when jointly scaling parallel and sequential test-time compute with a long reasoning R1 model. More broadly, RL$^V$ instantiates the principle of co-training for test-time scaling: jointly optimizing for task performance and a capability useful at inference, using data that RL training already produces. △ Less

Submitted 12 April, 2026; v1 submitted 7 May, 2025; originally announced May 2025.

DTFSal: Audio-Visual Dynamic Token Fusion for Video Saliency Prediction

Authors: Kiana Hooshanfar, Alireza Hosseini, Ahmad Kalhor, Babak Nadjar Araabi

Abstract: Audio-visual saliency prediction aims to mimic human visual attention by identifying salient regions in videos through the integration of both visual and auditory information. Although visual-only approaches have significantly advanced, effectively incorporating auditory cues remains challenging due to complex spatio-temporal interactions and high computational demands. To address these challenges… ▽ More Audio-visual saliency prediction aims to mimic human visual attention by identifying salient regions in videos through the integration of both visual and auditory information. Although visual-only approaches have significantly advanced, effectively incorporating auditory cues remains challenging due to complex spatio-temporal interactions and high computational demands. To address these challenges, we propose Dynamic Token Fusion Saliency (DFTSal), a novel audio-visual saliency prediction framework designed to balance accuracy with computational efficiency. Our approach features a multi-scale visual encoder equipped with two novel modules: the Learnable Token Enhancement Block (LTEB), which adaptively weights tokens to emphasize crucial saliency cues, and the Dynamic Learnable Token Fusion Block (DLTFB), which employs a shifting operation to reorganize and merge features, effectively capturing long-range dependencies and detailed spatial information. In parallel, an audio branch processes raw audio signals to extract meaningful auditory features. Both visual and audio features are integrated using our Adaptive Multimodal Fusion Block (AMFB), which employs local, global, and adaptive fusion streams for precise cross-modal fusion. The resulting fused features are processed by a hierarchical multi-decoder structure, producing accurate saliency maps. Extensive evaluations on six audio-visual benchmarks demonstrate that DFTSal achieves SOTA performance while maintaining computational efficiency. △ Less

Submitted 16 April, 2025; v1 submitted 14 April, 2025; originally announced April 2025.

Real-Time Roadway Obstacle Detection for Electric Scooters Using Deep Learning and Multi-Sensor Fusion

Authors: Zeyang Zheng, Arman Hosseini, Dong Chen, Omid Shoghli, Arsalan Heydarian

Abstract: The increasing adoption of electric scooters (e-scooters) in urban areas has coincided with a rise in traffic accidents and injuries, largely due to their small wheels, lack of suspension, and sensitivity to uneven surfaces. While deep learning-based object detection has been widely used to improve automobile safety, its application for e-scooter obstacle detection remains unexplored. This study i… ▽ More The increasing adoption of electric scooters (e-scooters) in urban areas has coincided with a rise in traffic accidents and injuries, largely due to their small wheels, lack of suspension, and sensitivity to uneven surfaces. While deep learning-based object detection has been widely used to improve automobile safety, its application for e-scooter obstacle detection remains unexplored. This study introduces a novel ground obstacle detection system for e-scooters, integrating an RGB camera, and a depth camera to enhance real-time road hazard detection. Additionally, the Inertial Measurement Unit (IMU) measures linear vertical acceleration to identify surface vibrations, guiding the selection of six obstacle categories: tree branches, manhole covers, potholes, pine cones, non-directional cracks, and truncated domes. All sensors, including the RGB camera, depth camera, and IMU, are integrated within the Intel RealSense Camera D435i. A deep learning model powered by YOLO detects road hazards and utilizes depth data to estimate obstacle proximity. Evaluated on the seven hours of naturalistic riding dataset, the system achieves a high mean average precision (mAP) of 0.827 and demonstrates excellent real-time performance. This approach provides an effective solution to enhance e-scooter safety through advanced computer vision and data fusion. The dataset is accessible at https://zenodo.org/records/14583718, and the project code is hosted on https://github.com/Zeyang-Zheng/Real-Time-Roadway-Obstacle-Detection-for-Electric-Scooters. △ Less

Submitted 4 April, 2025; originally announced April 2025.

Comments: Accepted at ASCE International Conference on Computing in Civil Engineering (i3ce)

When To Solve, When To Verify: Compute-Optimal Problem Solving and Generative Verification for LLM Reasoning

Authors: Nishad Singhi, Hritik Bansal, Arian Hosseini, Aditya Grover, Kai-Wei Chang, Marcus Rohrbach, Anna Rohrbach

Abstract: Scaling test-time compute has emerged as a key strategy for enhancing the reasoning capabilities of large language models (LLMs), particularly in tasks like mathematical problem-solving. A traditional approach, Self-Consistency (SC), generates multiple solutions to a problem and selects the most common answer via majority voting. Another common method involves scoring each solution with a reward m… ▽ More Scaling test-time compute has emerged as a key strategy for enhancing the reasoning capabilities of large language models (LLMs), particularly in tasks like mathematical problem-solving. A traditional approach, Self-Consistency (SC), generates multiple solutions to a problem and selects the most common answer via majority voting. Another common method involves scoring each solution with a reward model (verifier) and choosing the best one. Recent advancements in Generative Reward Models (GenRM) reframe verification as a next-token prediction task, enabling inference-time scaling along a new axis. Specifically, GenRM generates multiple verification chains-of-thought to score each solution. Under a limited inference budget, this introduces a fundamental trade-off: should you spend the budget on scaling solutions via SC or generate fewer solutions and allocate compute to verification via GenRM? To address this, we evaluate GenRM against SC under a fixed inference budget. Interestingly, we find that SC is more compute-efficient than GenRM for most practical inference budgets across diverse models and datasets. For instance, GenRM first matches SC after consuming up to 8x the inference compute and requires significantly more compute to outperform it. Furthermore, we derive inference scaling laws for the GenRM paradigm, revealing that compute-optimal inference favors scaling solution generation more aggressively than scaling the number of verifications. Our work provides practical guidance on optimizing test-time scaling by balancing solution generation and verification. The code is available at https://github.com/nishadsinghi/sc-genrm-scaling. △ Less

Submitted 19 October, 2025; v1 submitted 1 April, 2025; originally announced April 2025.

Comments: COLM 2025

Adoption of AI-Assisted E-Scooters: The Role of Perceived Trust, Safety, and Demographic Drivers

Authors: Amit Kumar, Arman Hosseini, Arghavan Azarbayjani, Arsalan Heydarian, Omidreza Shoghli

Abstract: E-scooters have become a more dominant mode of transport in recent years. However, the rise in their usage has been accompanied by an increase in injuries, affecting the trust and perceived safety of both users and non-users. Artificial intelligence (AI), as a cutting-edge and widely applied technology, has demonstrated potential to enhance transportation safety, particularly in driver assistance… ▽ More E-scooters have become a more dominant mode of transport in recent years. However, the rise in their usage has been accompanied by an increase in injuries, affecting the trust and perceived safety of both users and non-users. Artificial intelligence (AI), as a cutting-edge and widely applied technology, has demonstrated potential to enhance transportation safety, particularly in driver assistance systems. The integration of AI into e-scooters presents a promising approach to addressing these safety concerns. This study aims to explore the factors influencing individuals willingness to use AI-assisted e-scooters. Data were collected using a structured questionnaire, capturing responses from 405 participants. The questionnaire gathered information on demographic characteristics, micromobility usage frequency, road users' perception of safety around e-scooters, perceptions of safety in AI-enabled technology, trust in AI-enabled e-scooters, and involvement in e-scooter crash incidents. To examine the impact of demographic factors on participants' preferences between AI-assisted and regular e-scooters, decision tree analysis is employed, indicating that ethnicity, income, and age significantly influence preferences. To analyze the impact of other factors on the willingness to use AI-enabled e-scooters, a full-scale Structural Equation Model (SEM) is applied, revealing that the perception of safety in AI enabled technology and the level of trust in AI-enabled e-scooters are the strongest predictors. △ Less

Submitted 7 February, 2025; originally announced February 2025.

Multi-Agent Deep Q-Network with Layer-based Communication Channel for Autonomous Internal Logistics Vehicle Scheduling in Smart Manufacturing

Authors: Mohammad Feizabadi, Arman Hosseini, Zakaria Yahouni

Abstract: In smart manufacturing, scheduling autonomous internal logistic vehicles is crucial for optimizing operational efficiency. This paper proposes a multi-agent deep Q-network (MADQN) with a layer-based communication channel (LBCC) to address this challenge. The main goals are to minimize total job tardiness, reduce the number of tardy jobs, and lower vehicle energy consumption. The method is evaluate… ▽ More In smart manufacturing, scheduling autonomous internal logistic vehicles is crucial for optimizing operational efficiency. This paper proposes a multi-agent deep Q-network (MADQN) with a layer-based communication channel (LBCC) to address this challenge. The main goals are to minimize total job tardiness, reduce the number of tardy jobs, and lower vehicle energy consumption. The method is evaluated against nine well-known scheduling heuristics, demonstrating its effectiveness in handling dynamic job shop behaviors like job arrivals and workstation unavailabilities. The approach also proves scalable, maintaining performance across different layouts and larger problem instances, highlighting the robustness and adaptability of MADQN with LBCC in smart manufacturing. △ Less

Submitted 1 November, 2024; originally announced November 2024.

Comments: Accepted for the 5th IFAC/INSTICC INTERNATIONAL CONFERENCE ON INNOVATIVE INTELLIGENT INDUSTRIAL PRODUCTION AND LOGISTICS

Asynchronous RLHF: Faster and More Efficient Off-Policy RL for Language Models

Authors: Michael Noukhovitch, Shengyi Huang, Sophie Xhonneux, Arian Hosseini, Rishabh Agarwal, Aaron Courville

Abstract: The dominant paradigm for RLHF is online and on-policy RL: synchronously generating from the large language model (LLM) policy, labelling with a reward model, and learning using feedback on the LLM's own outputs. While performant, this paradigm is computationally inefficient. Inspired by classical deep RL literature, we propose separating generation and learning in RLHF. This enables asynchronous… ▽ More The dominant paradigm for RLHF is online and on-policy RL: synchronously generating from the large language model (LLM) policy, labelling with a reward model, and learning using feedback on the LLM's own outputs. While performant, this paradigm is computationally inefficient. Inspired by classical deep RL literature, we propose separating generation and learning in RLHF. This enables asynchronous generation of new samples while simultaneously training on old samples, leading to faster training and more compute-optimal scaling. However, asynchronous training relies on an underexplored regime, online but off-policy RLHF: learning on samples from previous iterations of our model which give a worse training signal. We tackle the fundamental challenge in this regime: how much off-policyness can we tolerate for asynchronous training to speed up learning but maintain performance? Among several RLHF algorithms we test, online DPO is found to be most robust to off-policy data, and robustness increases with the scale of the policy model. We study further compute optimizations for asynchronous RLHF but find that they come at a performance cost, giving rise to a trade-off. We verify the scalability of asynchronous RLHF by training a general-purpose chatbot from LLaMA 3.1 8B on an instruction-following task ~40% faster than a synchronous run while matching final performance. Finally, we extend our results to math and reasoning to demonstrate asynchronous RL can finetune Rho 1B on GSM8k ~70% faster while matching synchronous accuracy. △ Less

Submitted 26 April, 2025; v1 submitted 23 October, 2024; originally announced October 2024.

Comments: accepted at ICLR 2025, code at https://github.com/mnoukhov/async_rlhf, integrated into the open-instruct library https://github.com/allenai/open-instruct

Not All LLM Reasoners Are Created Equal

Authors: Arian Hosseini, Alessandro Sordoni, Daniel Toyama, Aaron Courville, Rishabh Agarwal

Abstract: We study the depth of grade-school math (GSM) problem-solving capabilities of LLMs. To this end, we evaluate their performance on pairs of existing math word problems together so that the answer to the second problem depends on correctly answering the first problem. Our findings reveal a significant reasoning gap in most LLMs, that is performance difference between solving the compositional pairs… ▽ More We study the depth of grade-school math (GSM) problem-solving capabilities of LLMs. To this end, we evaluate their performance on pairs of existing math word problems together so that the answer to the second problem depends on correctly answering the first problem. Our findings reveal a significant reasoning gap in most LLMs, that is performance difference between solving the compositional pairs and solving each question independently. This gap is more pronounced in smaller, more cost-efficient, and math-specialized models. Moreover, instruction-tuning recipes and code generation have varying effects across LLM sizes, while finetuning on GSM can lead to task overfitting. Our analysis indicates that large reasoning gaps are not because of test-set leakage, but due to distraction from additional context and poor second-hop reasoning. Overall, LLMs exhibit systematic differences in their reasoning abilities, despite what their performance on standard benchmarks indicates. △ Less

Submitted 2 October, 2024; originally announced October 2024.

Constructing an Interpretable Deep Denoiser by Unrolling Graph Laplacian Regularizer

Authors: Seyed Alireza Hosseini, Tam Thuc Do, Gene Cheung, Yuichi Tanaka

Abstract: An image denoiser can be used for a wide range of restoration problems via the Plug-and-Play (PnP) architecture. In this paper, we propose a general framework to build an interpretable graph-based deep denoiser (GDD) by unrolling a solution to a maximum a posteriori (MAP) problem equipped with a graph Laplacian regularizer (GLR) as signal prior. Leveraging a recent theorem showing that any (pseudo… ▽ More An image denoiser can be used for a wide range of restoration problems via the Plug-and-Play (PnP) architecture. In this paper, we propose a general framework to build an interpretable graph-based deep denoiser (GDD) by unrolling a solution to a maximum a posteriori (MAP) problem equipped with a graph Laplacian regularizer (GLR) as signal prior. Leveraging a recent theorem showing that any (pseudo-)linear denoiser $\boldsymbol Ψ$, under mild conditions, can be mapped to a solution of a MAP denoising problem regularized using GLR, we first initialize a graph Laplacian matrix $\mathbf L$ via truncated Taylor Series Expansion (TSE) of $\boldsymbol Ψ^{-1}$. Then, we compute the MAP linear system solution by unrolling iterations of the conjugate gradient (CG) algorithm into a sequence of neural layers as a feed-forward network -- one that is amenable to parameter tuning. The resulting GDD network is "graph-interpretable", low in parameter count, and easy to initialize thanks to $\mathbf L$ derived from a known well-performing denoiser $\boldsymbol Ψ$. Experimental results show that GDD achieves competitive image denoising performance compared to competitors, but employing far fewer parameters, and is more robust to covariate shift. △ Less

Submitted 10 September, 2024; originally announced September 2024.

Smaller, Weaker, Yet Better: Training LLM Reasoners via Compute-Optimal Sampling

Authors: Hritik Bansal, Arian Hosseini, Rishabh Agarwal, Vinh Q. Tran, Mehran Kazemi

Abstract: Training on high-quality synthetic data from strong language models (LMs) is a common strategy to improve the reasoning performance of LMs. In this work, we revisit whether this strategy is compute-optimal under a fixed inference budget (e.g., FLOPs). To do so, we investigate the trade-offs between generating synthetic data using a stronger but more expensive (SE) model versus a weaker but cheaper… ▽ More Training on high-quality synthetic data from strong language models (LMs) is a common strategy to improve the reasoning performance of LMs. In this work, we revisit whether this strategy is compute-optimal under a fixed inference budget (e.g., FLOPs). To do so, we investigate the trade-offs between generating synthetic data using a stronger but more expensive (SE) model versus a weaker but cheaper (WC) model. We evaluate the generated data across three key metrics: coverage, diversity, and false positive rate, and show that the data from WC models may have higher coverage and diversity, but also exhibit higher false positive rates. We then finetune LMs on data from SE and WC models in different settings: knowledge distillation, self-improvement, and a novel weak-to-strong improvement setup where a weaker LM teaches reasoning to a stronger LM. Our findings reveal that models finetuned on WC-generated data consistently outperform those trained on SE-generated data across multiple benchmarks and multiple choices of WC and SE models. These results challenge the prevailing practice of relying on SE models for synthetic data generation, suggesting that WC may be the compute-optimal approach for training advanced LM reasoners. △ Less

Submitted 7 October, 2024; v1 submitted 29 August, 2024; originally announced August 2024.

Generative Verifiers: Reward Modeling as Next-Token Prediction

Authors: Lunjun Zhang, Arian Hosseini, Hritik Bansal, Mehran Kazemi, Aviral Kumar, Rishabh Agarwal

Abstract: Verifiers or reward models are often used to enhance the reasoning performance of large language models (LLMs). A common approach is the Best-of-N method, where N candidate solutions generated by the LLM are ranked by a verifier, and the best one is selected. While LLM-based verifiers are typically trained as discriminative classifiers to score solutions, they do not utilize the text generation ca… ▽ More Verifiers or reward models are often used to enhance the reasoning performance of large language models (LLMs). A common approach is the Best-of-N method, where N candidate solutions generated by the LLM are ranked by a verifier, and the best one is selected. While LLM-based verifiers are typically trained as discriminative classifiers to score solutions, they do not utilize the text generation capabilities of pretrained LLMs. To overcome this limitation, we instead propose training verifiers using the ubiquitous next-token prediction objective, jointly on verification and solution generation. Compared to standard verifiers, such generative verifiers (GenRM) can benefit from several advantages of LLMs: they integrate seamlessly with instruction tuning, enable chain-of-thought reasoning, and can utilize additional test-time compute via majority voting for better verification. We demonstrate that GenRM outperforms discriminative, DPO verifiers, and LLM-as-a-Judge, resulting in large performance gains with Best-of-N, namely 5% $\rightarrow$ 45.3% on algorithmic tasks and 73% $\rightarrow$ 93.4% on GSM8K. In easy-to-hard generalization settings, we observe improvements of 28% $\rightarrow$ 44.6% on MATH, and 37.9% $\rightarrow$ 53.5% on MMLU abstract algebra. Furthermore, we find that training GenRM with synthetic verification rationales is sufficient to pick out subtle errors on math problems. Finally, we demonstrate that GenRM scales favorably with model size and test-time compute. △ Less

Submitted 22 February, 2025; v1 submitted 27 August, 2024; originally announced August 2024.

Comments: ICLR 2025

Leveraging LLM Reasoning Enhances Personalized Recommender Systems

Authors: Alicia Y. Tsai, Adam Kraft, Long Jin, Chenwei Cai, Anahita Hosseini, Taibai Xu, Zemin Zhang, Lichan Hong, Ed H. Chi, Xinyang Yi

Abstract: Recent advancements have showcased the potential of Large Language Models (LLMs) in executing reasoning tasks, particularly facilitated by Chain-of-Thought (CoT) prompting. While tasks like arithmetic reasoning involve clear, definitive answers and logical chains of thought, the application of LLM reasoning in recommendation systems (RecSys) presents a distinct challenge. RecSys tasks revolve arou… ▽ More Recent advancements have showcased the potential of Large Language Models (LLMs) in executing reasoning tasks, particularly facilitated by Chain-of-Thought (CoT) prompting. While tasks like arithmetic reasoning involve clear, definitive answers and logical chains of thought, the application of LLM reasoning in recommendation systems (RecSys) presents a distinct challenge. RecSys tasks revolve around subjectivity and personalized preferences, an under-explored domain in utilizing LLMs' reasoning capabilities. Our study explores several aspects to better understand reasoning for RecSys and demonstrate how task quality improves by utilizing LLM reasoning in both zero-shot and finetuning settings. Additionally, we propose RecSAVER (Recommender Systems Automatic Verification and Evaluation of Reasoning) to automatically assess the quality of LLM reasoning responses without the requirement of curated gold references or human raters. We show that our framework aligns with real human judgment on the coherence and faithfulness of reasoning responses. Overall, our work shows that incorporating reasoning into RecSys can improve personalized tasks, paving the way for further advancements in recommender system methodologies. △ Less

Submitted 22 July, 2024; originally announced August 2024.

Comments: To be published at ACL 2024

Impact of Road Infrastructure and Traffic Scenarios on E-scooterists' Riding and Gaze Behavior

Authors: Dong Chen, Arman Hosseini, Arik Smith, Zeyang Zheng, David Xiang, Arsalan Heydarian, Omid Shoghli, Bradford Campbell

Abstract: The growing adoption of e-scooters has raised significant safety concerns, particularly due to a surge in injuries and fatalities. This study explores the relationship between road infrastructure, traffic scenarios, and e-scooterists' riding and gaze behaviors to improve road safety and user experience. A naturalistic study was conducted using instrumented e-scooters, capturing gaze patterns, fixa… ▽ More The growing adoption of e-scooters has raised significant safety concerns, particularly due to a surge in injuries and fatalities. This study explores the relationship between road infrastructure, traffic scenarios, and e-scooterists' riding and gaze behaviors to improve road safety and user experience. A naturalistic study was conducted using instrumented e-scooters, capturing gaze patterns, fixation metrics, and head movement data across various road layouts and traffic scenarios. Key findings reveal that bike lanes offer a stable environment with reduced horizontal head movement and focused attention on the road, while shared roads and sidewalks lead to more dispersed gaze and increased head movement, indicating higher uncertainty and complexity. Interactions with other road users, such as navigating intersections, passing buses, riding near cars, and descending on downhill paths, demand greater cognitive load. Intersections require heightened visual focus and spatial awareness, reflected in increased horizontal eye and head movements. Interactions with vehicles prioritize visual scanning over head movement to maintain stability and avoid collisions, while high-speed and downhill riding demand focused attention on obstacles and the road surface. The results provide insights into e-scooter riders' behavior and physiological response analysis, paving the way for safer riding experiences and improved understanding of their needs. △ Less

Submitted 16 March, 2025; v1 submitted 5 May, 2024; originally announced July 2024.

Comments: 12 pages, 10 figures

Journal ref: International Conference on Transportation & Development (ICTD 2025)

SUM: Saliency Unification through Mamba for Visual Attention Modeling

Authors: Alireza Hosseini, Amirhossein Kazerouni, Saeed Akhavan, Michael Brudno, Babak Taati

Abstract: Visual attention modeling, important for interpreting and prioritizing visual stimuli, plays a significant role in applications such as marketing, multimedia, and robotics. Traditional saliency prediction models, especially those based on Convolutional Neural Networks (CNNs) or Transformers, achieve notable success by leveraging large-scale annotated datasets. However, the current state-of-the-art… ▽ More Visual attention modeling, important for interpreting and prioritizing visual stimuli, plays a significant role in applications such as marketing, multimedia, and robotics. Traditional saliency prediction models, especially those based on Convolutional Neural Networks (CNNs) or Transformers, achieve notable success by leveraging large-scale annotated datasets. However, the current state-of-the-art (SOTA) models that use Transformers are computationally expensive. Additionally, separate models are often required for each image type, lacking a unified approach. In this paper, we propose Saliency Unification through Mamba (SUM), a novel approach that integrates the efficient long-range dependency modeling of Mamba with U-Net to provide a unified model for diverse image types. Using a novel Conditional Visual State Space (C-VSS) block, SUM dynamically adapts to various image types, including natural scenes, web pages, and commercial imagery, ensuring universal applicability across different data types. Our comprehensive evaluations across five benchmarks demonstrate that SUM seamlessly adapts to different visual characteristics and consistently outperforms existing models. These results position SUM as a versatile and powerful tool for advancing visual attention modeling, offering a robust solution universally applicable across different types of visual content. △ Less

Submitted 9 September, 2024; v1 submitted 25 June, 2024; originally announced June 2024.

Comments: Accepted at IEEE/CVF WACV 2025

Interpretable Lightweight Transformer via Unrolling of Learned Graph Smoothness Priors

Authors: Tam Thuc Do, Parham Eftekhar, Seyed Alireza Hosseini, Gene Cheung, Philip Chou

Abstract: We build interpretable and lightweight transformer-like neural networks by unrolling iterative optimization algorithms that minimize graph smoothness priors -- the quadratic graph Laplacian regularizer (GLR) and the $\ell_1$-norm graph total variation (GTV) -- subject to an interpolation constraint. The crucial insight is that a normalized signal-dependent graph learning module amounts to a varian… ▽ More We build interpretable and lightweight transformer-like neural networks by unrolling iterative optimization algorithms that minimize graph smoothness priors -- the quadratic graph Laplacian regularizer (GLR) and the $\ell_1$-norm graph total variation (GTV) -- subject to an interpolation constraint. The crucial insight is that a normalized signal-dependent graph learning module amounts to a variant of the basic self-attention mechanism in conventional transformers. Unlike "black-box" transformers that require learning of large key, query and value matrices to compute scaled dot products as affinities and subsequent output embeddings, resulting in huge parameter sets, our unrolled networks employ shallow CNNs to learn low-dimensional features per node to establish pairwise Mahalanobis distances and construct sparse similarity graphs. At each layer, given a learned graph, the target interpolated signal is simply a low-pass filtered output derived from the minimization of an assumed graph smoothness prior, leading to a dramatic reduction in parameter count. Experiments for two image interpolation applications verify the restoration performance, parameter efficiency and robustness to covariate shift of our graph-based unrolled networks compared to conventional transformers. △ Less

Submitted 5 November, 2024; v1 submitted 6 June, 2024; originally announced June 2024.

Performance Evaluation of Real-Time Object Detection for Electric Scooters

Authors: Dong Chen, Arman Hosseini, Arik Smith, Amir Farzin Nikkhah, Arsalan Heydarian, Omid Shoghli, Bradford Campbell

Abstract: Electric scooters (e-scooters) have rapidly emerged as a popular mode of transportation in urban areas, yet they pose significant safety challenges. In the United States, the rise of e-scooters has been marked by a concerning increase in related injuries and fatalities. Recently, while deep-learning object detection holds paramount significance in autonomous vehicles to avoid potential collisions,… ▽ More Electric scooters (e-scooters) have rapidly emerged as a popular mode of transportation in urban areas, yet they pose significant safety challenges. In the United States, the rise of e-scooters has been marked by a concerning increase in related injuries and fatalities. Recently, while deep-learning object detection holds paramount significance in autonomous vehicles to avoid potential collisions, its application in the context of e-scooters remains relatively unexplored. This paper addresses this gap by assessing the effectiveness and efficiency of cutting-edge object detectors designed for e-scooters. To achieve this, the first comprehensive benchmark involving 22 state-of-the-art YOLO object detectors, including five versions (YOLOv3, YOLOv5, YOLOv6, YOLOv7, and YOLOv8), has been established for real-time traffic object detection using a self-collected dataset featuring e-scooters. The detection accuracy, measured in terms of mAP@0.5, ranges from 27.4% (YOLOv7-E6E) to 86.8% (YOLOv5s). All YOLO models, particularly YOLOv3-tiny, have displayed promising potential for real-time object detection in the context of e-scooters. Both the traffic scene dataset (https://zenodo.org/records/10578641) and software program codes (https://github.com/DongChen06/ScooterDet) for model benchmarking in this study are publicly available, which will not only improve e-scooter safety with advanced object detection but also lay the groundwork for tailored solutions, promising a safer and more sustainable urban micromobility landscape. △ Less

Submitted 5 May, 2024; originally announced May 2024.

Comments: 10 pages, 3 figures

The N+ Implementation Details of RLHF with PPO: A Case Study on TL;DR Summarization

Authors: Shengyi Huang, Michael Noukhovitch, Arian Hosseini, Kashif Rasul, Weixun Wang, Lewis Tunstall

Abstract: This work is the first to openly reproduce the Reinforcement Learning from Human Feedback (RLHF) scaling behaviors reported in OpenAI's seminal TL;DR summarization work. We create an RLHF pipeline from scratch, enumerate over 20 key implementation details, and share key insights during the reproduction. Our RLHF-trained Pythia models demonstrate significant gains in response quality that scale wit… ▽ More This work is the first to openly reproduce the Reinforcement Learning from Human Feedback (RLHF) scaling behaviors reported in OpenAI's seminal TL;DR summarization work. We create an RLHF pipeline from scratch, enumerate over 20 key implementation details, and share key insights during the reproduction. Our RLHF-trained Pythia models demonstrate significant gains in response quality that scale with model size, with our 2.8B, 6.9B models outperforming OpenAI's released 1.3B checkpoint. We publicly release the trained model checkpoints and code to facilitate further research and accelerate progress in the field (\url{https://github.com/vwxyzjn/summarize_from_feedback_details}). △ Less

Submitted 23 March, 2024; originally announced March 2024.

Brand Visibility in Packaging: A Deep Learning Approach for Logo Detection, Saliency-Map Prediction, and Logo Placement Analysis

Authors: Alireza Hosseini, Kiana Hooshanfar, Pouria Omrani, Reza Toosi, Ramin Toosi, Zahra Ebrahimian, Mohammad Ali Akhaee

Abstract: In the highly competitive area of product marketing, the visibility of brand logos on packaging plays a crucial role in shaping consumer perception, directly influencing the success of the product. This paper introduces a comprehensive framework to measure the brand logo's attention on a packaging design. The proposed method consists of three steps. The first step leverages YOLOv8 for precise logo… ▽ More In the highly competitive area of product marketing, the visibility of brand logos on packaging plays a crucial role in shaping consumer perception, directly influencing the success of the product. This paper introduces a comprehensive framework to measure the brand logo's attention on a packaging design. The proposed method consists of three steps. The first step leverages YOLOv8 for precise logo detection across prominent datasets, FoodLogoDet-1500 and LogoDet-3K. The second step involves modeling the user's visual attention with a novel saliency prediction model tailored for the packaging context. The proposed saliency model combines the visual elements with text maps employing a transformers-based architecture to predict user attention maps. In the third step, by integrating logo detection with a saliency map generation, the framework provides a comprehensive brand attention score. The effectiveness of the proposed method is assessed module by module, ensuring a thorough evaluation of each component. Comparing logo detection and saliency map prediction with state-of-the-art models shows the superiority of the proposed methods. To investigate the robustness of the proposed brand attention score, we collected a unique dataset to examine previous psychophysical hypotheses related to brand visibility. the results show that the brand attention score is in line with all previous studies. Also, we introduced seven new hypotheses to check the impact of position, orientation, presence of person, and other visual elements on brand attention. This research marks a significant stride in the intersection of cognitive psychology, computer vision, and marketing, paving the way for advanced, consumer-centric packaging designs. △ Less

Submitted 4 March, 2024; originally announced March 2024.

V-STaR: Training Verifiers for Self-Taught Reasoners

Authors: Arian Hosseini, Xingdi Yuan, Nikolay Malkin, Aaron Courville, Alessandro Sordoni, Rishabh Agarwal

Abstract: Common self-improvement approaches for large language models (LLMs), such as STaR, iteratively fine-tune LLMs on self-generated solutions to improve their problem-solving ability. However, these approaches discard the large amounts of incorrect solutions generated during this process, potentially neglecting valuable information in such solutions. To address this shortcoming, we propose V-STaR that… ▽ More Common self-improvement approaches for large language models (LLMs), such as STaR, iteratively fine-tune LLMs on self-generated solutions to improve their problem-solving ability. However, these approaches discard the large amounts of incorrect solutions generated during this process, potentially neglecting valuable information in such solutions. To address this shortcoming, we propose V-STaR that utilizes both the correct and incorrect solutions generated during the self-improvement process to train a verifier using DPO that judges correctness of model-generated solutions. This verifier is used at inference time to select one solution among many candidate solutions. Running V-STaR for multiple iterations results in progressively better reasoners and verifiers, delivering a 4% to 17% test accuracy improvement over existing self-improvement and verification approaches on common code generation and math reasoning benchmarks with LLaMA2 models. △ Less

Submitted 13 August, 2024; v1 submitted 9 February, 2024; originally announced February 2024.

Large language models implicitly learn to straighten neural sentence trajectories to construct a predictive representation of natural language

Authors: Eghbal A. Hosseini, Evelina Fedorenko

Abstract: Predicting upcoming events is critical to our ability to interact with our environment. Transformer models, trained on next-word prediction, appear to construct representations of linguistic input that can support diverse downstream tasks. But how does a predictive objective shape such representations? Inspired by recent work in vision (Henaff et al., 2019), we test a hypothesis about predictive r… ▽ More Predicting upcoming events is critical to our ability to interact with our environment. Transformer models, trained on next-word prediction, appear to construct representations of linguistic input that can support diverse downstream tasks. But how does a predictive objective shape such representations? Inspired by recent work in vision (Henaff et al., 2019), we test a hypothesis about predictive representations of autoregressive transformers. In particular, we test whether the neural trajectory of a sentence becomes progressively straighter as it passes through the network layers. The key insight is that straighter trajectories should facilitate prediction via linear extrapolation. We quantify straightness using a 1-dimensional curvature metric, and present four findings in support of the trajectory straightening hypothesis: i) In trained models, the curvature decreases from the early to the deeper layers of the network. ii) Models that perform better on the next-word prediction objective exhibit greater decreases in curvature, suggesting that this improved ability to straighten sentence trajectories may be the driver of better language modeling performance. iii) Given the same linguistic context, the sequences that are generated by the model have lower curvature than the actual continuations observed in a language corpus, suggesting that the model favors straighter trajectories for making predictions. iv) A consistent relationship holds between the average curvature and the average surprisal of sentences in the deep model layers, such that sentences with straighter trajectories also have lower surprisal. Importantly, untrained models do not exhibit these behaviors. In tandem, these results support the trajectory straightening hypothesis and provide a possible mechanism for how the geometry of the internal representations of autoregressive models supports next word prediction. △ Less

Submitted 5 November, 2023; originally announced November 2023.

Comments: 37th Conference on Neural Information Processing Systems (NeurIPS 2023). 20 pages, 5 main figures, 7 supplementary figures

INCODE: Implicit Neural Conditioning with Prior Knowledge Embeddings

Authors: Amirhossein Kazerouni, Reza Azad, Alireza Hosseini, Dorit Merhof, Ulas Bagci

Abstract: Implicit Neural Representations (INRs) have revolutionized signal representation by leveraging neural networks to provide continuous and smooth representations of complex data. However, existing INRs face limitations in capturing fine-grained details, handling noise, and adapting to diverse signal types. To address these challenges, we introduce INCODE, a novel approach that enhances the control o… ▽ More Implicit Neural Representations (INRs) have revolutionized signal representation by leveraging neural networks to provide continuous and smooth representations of complex data. However, existing INRs face limitations in capturing fine-grained details, handling noise, and adapting to diverse signal types. To address these challenges, we introduce INCODE, a novel approach that enhances the control of the sinusoidal-based activation function in INRs using deep prior knowledge. INCODE comprises a harmonizer network and a composer network, where the harmonizer network dynamically adjusts key parameters of the activation function. Through a task-specific pre-trained model, INCODE adapts the task-specific parameters to optimize the representation process. Our approach not only excels in representation, but also extends its prowess to tackle complex tasks such as audio, image, and 3D shape reconstructions, as well as intricate challenges such as neural radiance fields (NeRFs), and inverse problems, including denoising, super-resolution, inpainting, and CT reconstruction. Through comprehensive experiments, INCODE demonstrates its superiority in terms of robustness, accuracy, quality, and convergence rate, broadening the scope of signal representation. Please visit the project's website for details on the proposed method and access to the code. △ Less

Submitted 28 October, 2023; originally announced October 2023.

Comments: Accepted at WACV 2024 conference

Epsilon non-Greedy: A Bandit Approach for Unbiased Recommendation via Uniform Data

Authors: S. M. F. Sani, Seyed Abbas Hosseini, Hamid R. Rabiee

Abstract: Often, recommendation systems employ continuous training, leading to a self-feedback loop bias in which the system becomes biased toward its previous recommendations. Recent studies have attempted to mitigate this bias by collecting small amounts of unbiased data. While these studies have successfully developed less biased models, they ignore the crucial fact that the recommendations generated by… ▽ More Often, recommendation systems employ continuous training, leading to a self-feedback loop bias in which the system becomes biased toward its previous recommendations. Recent studies have attempted to mitigate this bias by collecting small amounts of unbiased data. While these studies have successfully developed less biased models, they ignore the crucial fact that the recommendations generated by the model serve as the training data for subsequent training sessions. To address this issue, we propose a framework that learns an unbiased estimator using a small amount of uniformly collected data and focuses on generating improved training data for subsequent training iterations. To accomplish this, we view recommendation as a contextual multi-arm bandit problem and emphasize on exploring items that the model has a limited understanding of. We introduce a new offline sequential training schema that simulates real-world continuous training scenarios in recommendation systems, offering a more appropriate framework for studying self-feedback bias. We demonstrate the superiority of our model over state-of-the-art debiasing methods by conducting extensive experiments using the proposed training schema. △ Less

Submitted 7 October, 2023; originally announced October 2023.

Intelligent Traffic Control with Smart Speed Bumps

Authors: Melvin Mokhtari, Amirreza Hosseini, Alireza Habibi, Adel Karshenas, Ali Amoomahdi

Abstract: Traffic congestion and safety continue to pose significant challenges in urban environments. In this paper, we introduce the Smart Speed Bump (SSBump), a novel traffic calming solution that leverages the Internet of Things (IoT) and innovative non-Newtonian fluid materials to enhance road safety, optimize emergency response times, and improve the overall driving experience. The SSBump uses IoT sen… ▽ More Traffic congestion and safety continue to pose significant challenges in urban environments. In this paper, we introduce the Smart Speed Bump (SSBump), a novel traffic calming solution that leverages the Internet of Things (IoT) and innovative non-Newtonian fluid materials to enhance road safety, optimize emergency response times, and improve the overall driving experience. The SSBump uses IoT sensors to detect and communicate with emergency vehicles, reducing response times by temporarily deflating. These sensors also analyze traffic patterns and inform data-driven decisions. Additionally, the SSBump uses an Oobleck mixture that adapts its behavior based on the velocity of approaching vehicles, resulting in a safer and more comfortable experience for drivers. This study commences with an overview of the prevalent traffic congestion, followed by a discussion on various available options in this domain. Subsequently, the paper explores the advantages of smart speed bumps and their operational mechanisms. Finally, it presents a comprehensive analysis of the results, its challenges, and the prospects of the work. The findings of this research demonstrate the potential of the SSBump system to revolutionize traffic control, emergency response time, and the driving experience in smart cities, making it a game-changing innovation for advanced transportation systems. △ Less

Submitted 29 September, 2023; v1 submitted 1 July, 2023; originally announced July 2023.

Comments: 7 pages, 5 figures

Joint Prompt Optimization of Stacked LLMs using Variational Inference

Authors: Alessandro Sordoni, Xingdi Yuan, Marc-Alexandre Côté, Matheus Pereira, Adam Trischler, Ziang Xiao, Arian Hosseini, Friederike Niedtner, Nicolas Le Roux

Abstract: Large language models (LLMs) can be seen as atomic units of computation mapping sequences to a distribution over sequences. Thus, they can be seen as stochastic language layers in a language network, where the learnable parameters are the natural language prompts at each layer. By stacking two such layers and feeding the output of one layer to the next, we obtain a Deep Language Network (DLN). We… ▽ More Large language models (LLMs) can be seen as atomic units of computation mapping sequences to a distribution over sequences. Thus, they can be seen as stochastic language layers in a language network, where the learnable parameters are the natural language prompts at each layer. By stacking two such layers and feeding the output of one layer to the next, we obtain a Deep Language Network (DLN). We first show how to effectively perform prompt optimization for a 1-Layer language network (DLN-1). Then, we present an extension that applies to 2-layer DLNs (DLN-2), where two prompts must be learned. The key idea is to consider the output of the first layer as a latent variable, which requires inference, and prompts to be learned as the parameters of the generative distribution. We first test the effectiveness of DLN-1 in multiple reasoning and natural language understanding tasks. Then, we show that DLN-2 can reach higher performance than a single layer, showing promise that we might reach comparable performance to GPT-4, even when each LLM in the network is smaller and less powerful. △ Less

Submitted 4 December, 2023; v1 submitted 21 June, 2023; originally announced June 2023.

Comments: NeurIPS 2023

Emotional Framing in the Spreading of False and True Claims

Authors: Akram Sadat Hosseini, Steffen Staab

Abstract: The explosive growth of online misinformation, such as false claims, has affected the social behavior of online users. In order to be persuasive and mislead the audience, false claims are made to trigger emotions in their audience. This paper contributes to understanding how misinformation in social media is shaped by investigating the emotional framing that authors of the claims try to create for… ▽ More The explosive growth of online misinformation, such as false claims, has affected the social behavior of online users. In order to be persuasive and mislead the audience, false claims are made to trigger emotions in their audience. This paper contributes to understanding how misinformation in social media is shaped by investigating the emotional framing that authors of the claims try to create for their audience. We investigate how, firstly, the existence of emotional framing in the claims depends on the topic and credibility of the claims. Secondly, we explore how emotionally framed content triggers emotional response posts by social media users, and how emotions expressed in claims and corresponding users' response posts affect their sharing behavior on social media. Analysis of four data sets covering different topics (politics, health, Syrian war, and COVID-19) reveals that authors shape their claims depending on the topic area to pass targeted emotions to their audience. By analysing responses to claims, we show that the credibility of the claim influences the distribution of emotions that the claim incites in its audience. Moreover, our analysis shows that emotions expressed in the claims are repeated in the users' responses. Finally, the analysis of users' sharing behavior shows that negative emotional framing such as anger, fear, and sadness of false claims leads to more interaction among users than positive emotions. This analysis also reveals that in the claims that trigger happy responses, true claims result in more sharing compared to false claims. △ Less

Submitted 29 March, 2023; originally announced March 2023.

On the Compositional Generalization Gap of In-Context Learning

Authors: Arian Hosseini, Ankit Vani, Dzmitry Bahdanau, Alessandro Sordoni, Aaron Courville

Abstract: Pretrained large generative language models have shown great performance on many tasks, but exhibit low compositional generalization abilities. Scaling such models has been shown to improve their performance on various NLP tasks even just by conditioning them on a few examples to solve the task without any fine-tuning (also known as in-context learning). In this work, we look at the gap between th… ▽ More Pretrained large generative language models have shown great performance on many tasks, but exhibit low compositional generalization abilities. Scaling such models has been shown to improve their performance on various NLP tasks even just by conditioning them on a few examples to solve the task without any fine-tuning (also known as in-context learning). In this work, we look at the gap between the in-distribution (ID) and out-of-distribution (OOD) performance of such models in semantic parsing tasks with in-context learning. In the ID settings, the demonstrations are from the same split (test or train) that the model is being evaluated on, and in the OOD settings, they are from the other split. We look at how the relative generalization gap of in-context learning evolves as models are scaled up. We evaluate four model families, OPT, BLOOM, CodeGen and Codex on three semantic parsing datasets, CFQ, SCAN and GeoQuery with different number of exemplars, and observe a trend of decreasing relative generalization gap as models are scaled up. △ Less

Submitted 15 November, 2022; originally announced November 2022.

The Interplay between Distribution Parameters and the Accuracy-Robustness Tradeoff in Classification

Authors: Alireza Mousavi Hosseini, Amir Mohammad Abouei, Mohammad Hossein Rohban

Abstract: Adversarial training tends to result in models that are less accurate on natural (unperturbed) examples compared to standard models. This can be attributed to either an algorithmic shortcoming or a fundamental property of the training data distribution, which admits different solutions for optimal standard and adversarial classifiers. In this work, we focus on the latter case under a binary Gaussi… ▽ More Adversarial training tends to result in models that are less accurate on natural (unperturbed) examples compared to standard models. This can be attributed to either an algorithmic shortcoming or a fundamental property of the training data distribution, which admits different solutions for optimal standard and adversarial classifiers. In this work, we focus on the latter case under a binary Gaussian mixture classification problem. Unlike earlier work, we aim to derive the natural accuracy gap between the optimal Bayes and adversarial classifiers, and study the effect of different distributional parameters, namely separation between class centroids, class proportions, and the covariance matrix, on the derived gap. We show that under certain conditions, the natural error of the optimal adversarial classifier, as well as the gap, are locally minimized when classes are balanced, contradicting the performance of the Bayes classifier where perfect balance induces the worst accuracy. Moreover, we show that with an $\ell_\infty$ bounded perturbation and an adversarial budget of $ε$, this gap is $Θ(ε^2)$ for the worst-case parameters, which for suitably small $ε$ indicates the theoretical possibility of achieving robust classifiers with near-perfect accuracy, which is rarely reflected in practical algorithms. △ Less

Submitted 1 July, 2021; originally announced July 2021.

Comments: Accepted for presentation in AML ICML workshop 2021

Understanding by Understanding Not: Modeling Negation in Language Models

Authors: Arian Hosseini, Siva Reddy, Dzmitry Bahdanau, R Devon Hjelm, Alessandro Sordoni, Aaron Courville

Abstract: Negation is a core construction in natural language. Despite being very successful on many tasks, state-of-the-art pre-trained language models often handle negation incorrectly. To improve language models in this regard, we propose to augment the language modeling objective with an unlikelihood objective that is based on negated generic sentences from a raw text corpus. By training BERT with the r… ▽ More Negation is a core construction in natural language. Despite being very successful on many tasks, state-of-the-art pre-trained language models often handle negation incorrectly. To improve language models in this regard, we propose to augment the language modeling objective with an unlikelihood objective that is based on negated generic sentences from a raw text corpus. By training BERT with the resulting combined objective we reduce the mean top~1 error rate to 4% on the negated LAMA dataset. We also see some improvements on the negated NLI benchmarks. △ Less

Submitted 7 May, 2021; originally announced May 2021.

Zero-Shot Self-Supervised Learning for MRI Reconstruction

Authors: Burhaneddin Yaman, Seyed Amir Hossein Hosseini, Mehmet Akçakaya

Abstract: Deep learning (DL) has emerged as a powerful tool for accelerated MRI reconstruction, but often necessitates a database of fully-sampled measurements for training. Recent self-supervised and unsupervised learning approaches enable training without fully-sampled data. However, a database of undersampled measurements may not be available in many scenarios, especially for scans involving contrast or… ▽ More Deep learning (DL) has emerged as a powerful tool for accelerated MRI reconstruction, but often necessitates a database of fully-sampled measurements for training. Recent self-supervised and unsupervised learning approaches enable training without fully-sampled data. However, a database of undersampled measurements may not be available in many scenarios, especially for scans involving contrast or translational acquisitions in development. Moreover, recent studies show that database-trained models may not generalize well when the unseen measurements differ in terms of sampling pattern, acceleration rate, SNR, image contrast, and anatomy. Such challenges necessitate a new methodology to enable subject-specific DL MRI reconstruction without external training datasets, since it is clinically imperative to provide high-quality reconstructions that can be used to identify lesions/disease for \emph{every individual}. In this work, we propose a zero-shot self-supervised learning approach to perform subject-specific accelerated DL MRI reconstruction to tackle these issues. The proposed approach partitions the available measurements from a single scan into three disjoint sets. Two of these sets are used to enforce data consistency and define loss during training for self-supervision, while the last set serves to self-validate, establishing an early stopping criterion. In the presence of models pre-trained on a database with different image characteristics, we show that the proposed approach can be combined with transfer learning for faster convergence time and reduced computational complexity. The code is available at \url{https://github.com/byaman14/ZS-SSL}. △ Less

Submitted 28 November, 2023; v1 submitted 15 February, 2021; originally announced February 2021.

Journal ref: International Conference on Learning Representations (ICLR), 2022

Improved Supervised Training of Physics-Guided Deep Learning Image Reconstruction with Multi-Masking

Authors: Burhaneddin Yaman, Seyed Amir Hossein Hosseini, Steen Moeller, Mehmet Akçakaya

Abstract: Physics-guided deep learning (PG-DL) via algorithm unrolling has received significant interest for improved image reconstruction, including MRI applications. These methods unroll an iterative optimization algorithm into a series of regularizer and data consistency units. The unrolled networks are typically trained end-to-end using a supervised approach. Current supervised PG-DL approaches use all… ▽ More Physics-guided deep learning (PG-DL) via algorithm unrolling has received significant interest for improved image reconstruction, including MRI applications. These methods unroll an iterative optimization algorithm into a series of regularizer and data consistency units. The unrolled networks are typically trained end-to-end using a supervised approach. Current supervised PG-DL approaches use all of the available sub-sampled measurements in their data consistency units. Thus, the network learns to fit the rest of the measurements. In this study, we propose to improve the performance and robustness of supervised training by utilizing randomness by retrospectively selecting only a subset of all the available measurements for data consistency units. The process is repeated multiple times using different random masks during training for further enhancement. Results on knee MRI show that the proposed multi-mask supervised PG-DL enhances reconstruction performance compared to conventional supervised PG-DL approaches. △ Less

Submitted 26 October, 2020; originally announced October 2020.

Journal ref: Proceedings of IEEE ICASSP, 2021

A Variational Auto-Encoder for Reservoir Monitoring

Authors: Kristian Gundersen, Seyyed A. Hosseini, Anna Oleynik, Guttorm Alendal

Abstract: Carbon dioxide Capture and Storage (CCS) is an important strategy in mitigating anthropogenic CO$_2$ emissions. In order for CCS to be successful, large quantities of CO$_2$ must be stored and the storage site conformance must be monitored. Here we present a deep learning method to reconstruct pressure fields and classify the flux out of the storage formation based on the pressure data from Above… ▽ More Carbon dioxide Capture and Storage (CCS) is an important strategy in mitigating anthropogenic CO$_2$ emissions. In order for CCS to be successful, large quantities of CO$_2$ must be stored and the storage site conformance must be monitored. Here we present a deep learning method to reconstruct pressure fields and classify the flux out of the storage formation based on the pressure data from Above Zone Monitoring Interval (AZMI) wells. The deep learning method is a version of a semi conditional variational auto-encoder tailored to solve two tasks: reconstruction of an incremental pressure field and leakage rate classification. The method, predictions and associated uncertainty estimates are illustrated on the synthetic data from a high-fidelity heterogeneous 2D numerical reservoir model, which was used to simulate subsurface CO$_2$ movement and pressure changes in the AZMI due to a CO$_2$ leakage. △ Less

Submitted 2 October, 2020; v1 submitted 23 September, 2020; originally announced September 2020.

Design and Implementation of a Maxi-Sized Mobile Robot (Karo) for Rescue Missions

Authors: Soheil Habibian, Mehdi Dadvar, Behzad Peykari, Alireza Hosseini, M. Hossein Salehzadeh, Alireza H. M. Hosseini, Farshid Najafi

Abstract: Rescue robots are expected to carry out reconnaissance and dexterity operations in unknown environments comprising unstructured obstacles. Although a wide variety of designs and implementations have been presented within the field of rescue robotics, embedding all mobility, dexterity, and reconnaissance capabilities in a single robot remains a challenging problem. This paper explains the design an… ▽ More Rescue robots are expected to carry out reconnaissance and dexterity operations in unknown environments comprising unstructured obstacles. Although a wide variety of designs and implementations have been presented within the field of rescue robotics, embedding all mobility, dexterity, and reconnaissance capabilities in a single robot remains a challenging problem. This paper explains the design and implementation of Karo, a mobile robot that exhibits a high degree of mobility at the side of maintaining required dexterity and exploration capabilities for urban search and rescue (USAR) missions. We first elicit the system requirements of a standard rescue robot from the frameworks of Rescue Robot League (RRL) of RoboCup and then, propose the conceptual design of Karo by drafting a locomotion and manipulation system. Considering that, this work presents comprehensive design processes along with detail mechanical design of the robot's platform and its 7-DOF manipulator. Further, we present the design and implementation of the command and control system by discussing the robot's power system, sensors, and hardware systems. In conjunction with this, we elucidate the way that Karo's software system and human-robot interface are implemented and employed. Furthermore, we undertake extensive evaluations of Karo's field performance to investigate whether the principal objective of this work has been satisfied. We demonstrate that Karo has effectively accomplished assigned standardized rescue operations by evaluating all aspects of its capabilities in both RRL's test suites and training suites of a fire department. Finally, the comprehensiveness of Karo's capabilities has been verified by drawing quantitative comparisons between Karo's performance and other leading robots participating in RRL. △ Less

Submitted 9 January, 2021; v1 submitted 23 July, 2020; originally announced August 2020.

Journal ref: Robomech J 8, 1 (2021)

Multi-Mask Self-Supervised Learning for Physics-Guided Neural Networks in Highly Accelerated MRI

Authors: Burhaneddin Yaman, Hongyi Gu, Seyed Amir Hossein Hosseini, Omer Burak Demirel, Steen Moeller, Jutta Ellermann, Kâmil Uğurbil, Mehmet Akçakaya

Abstract: Self-supervised learning has shown great promise due to its capability to train deep learning MRI reconstruction methods without fully-sampled data. Current self-supervised learning methods for physics-guided reconstruction networks split acquired undersampled data into two disjoint sets, where one is used for data consistency (DC) in the unrolled network and the other to define the training loss.… ▽ More Self-supervised learning has shown great promise due to its capability to train deep learning MRI reconstruction methods without fully-sampled data. Current self-supervised learning methods for physics-guided reconstruction networks split acquired undersampled data into two disjoint sets, where one is used for data consistency (DC) in the unrolled network and the other to define the training loss. In this study, we propose an improved self-supervised learning strategy that more efficiently uses the acquired data to train a physics-guided reconstruction network without a database of fully-sampled data. The proposed multi-mask self-supervised learning via data undersampling (SSDU) applies a hold-out masking operation on acquired measurements to split it into multiple pairs of disjoint sets for each training sample, while using one of these pairs for DC units and the other for defining loss, thereby more efficiently using the undersampled data. Multi-mask SSDU is applied on fully-sampled 3D knee and prospectively undersampled 3D brain MRI datasets, for various acceleration rates and patterns, and compared to CG-SENSE and single-mask SSDU DL-MRI, as well as supervised DL-MRI when fully-sampled data is available. Results on knee MRI show that the proposed multi-mask SSDU outperforms SSDU and performs closely with supervised DL-MRI. A clinical reader study further ranks the multi-mask SSDU higher than supervised DL-MRI in terms of SNR and aliasing artifacts. Results on brain MRI show that multi-mask SSDU achieves better reconstruction quality compared to SSDU. Reader study demonstrates that multi-mask SSDU at R=8 significantly improves reconstruction compared to single-mask SSDU at R=8, as well as CG-SENSE at R=2. △ Less

Submitted 8 June, 2022; v1 submitted 13 August, 2020; originally announced August 2020.

Journal ref: NMR in Biomedicine, 2022

Noise2Inpaint: Learning Referenceless Denoising by Inpainting Unrolling

Authors: Burhaneddin Yaman, Seyed Amir Hossein Hosseini, Mehmet Akçakaya

Abstract: Deep learning based image denoising methods have been recently popular due to their improved performance. Traditionally, these methods are trained in a supervised manner, requiring a set of noisy input and clean target image pairs. More recently, self-supervised approaches have been proposed to learn denoising from only noisy images. These methods assume that noise across pixels is statistically i… ▽ More Deep learning based image denoising methods have been recently popular due to their improved performance. Traditionally, these methods are trained in a supervised manner, requiring a set of noisy input and clean target image pairs. More recently, self-supervised approaches have been proposed to learn denoising from only noisy images. These methods assume that noise across pixels is statistically independent, and the underlying image pixels show spatial correlations across neighborhoods. These methods rely on a masking approach that divides the image pixels into two disjoint sets, where one is used as input to the network while the other is used to define the loss. However, these previous self-supervised approaches rely on a purely data-driven regularization neural network without explicitly taking the masking model into account. In this work, building on these self-supervised approaches, we introduce Noise2Inpaint (N2I), a training approach that recasts the denoising problem into a regularized image inpainting framework. This allows us to use an objective function, which can incorporate different statistical properties of the noise as needed. We use algorithm unrolling to unroll an iterative optimization for solving this objective function and train the unrolled network end-to-end. The training paradigm follows the masking approach from previous works, splitting the pixels into two disjoint sets. Importantly, one of these is now used to impose data fidelity in the unrolled network, while the other still defines the loss. We demonstrate that N2I performs successful denoising on real-world datasets, while better preserving details compared to its purely data-driven counterpart Noise2Self. △ Less

Submitted 19 November, 2020; v1 submitted 16 June, 2020; originally announced June 2020.