Showing 1–50 of 59 results for author: Ko, C

Evoflux: Inference-Time Evolution of Executable Tool Workflows for Compact Agents

Authors: Kushal Raj Bhandari, Ling Yue, Ching-Yun Ko, Dhaval Patel, Shaowu Pan, Pin-Yu Chen, Jianxi Gao

Abstract: Compact language models (LMs) reduce cost, latency, and deployment risk for tool agents. Yet MCP-style tool use requires more than isolated function calling: an agent must discover tools from live catalogs, satisfy schemas, preserve dependencies across intermediate outputs, and ground final responses in executed evidence. Small planners often generate plausible workflow graphs that fail under tool… ▽ More Compact language models (LMs) reduce cost, latency, and deployment risk for tool agents. Yet MCP-style tool use requires more than isolated function calling: an agent must discover tools from live catalogs, satisfy schemas, preserve dependencies across intermediate outputs, and ground final responses in executed evidence. Small planners often generate plausible workflow graphs that fail under tool resolution, parameter validation, dependency tracking, or execution. We argue that this failure mode is poorly handled by small-corpus distillation. A few hundred teacher traces can teach workflow format, but rarely cover the recovery behavior needed to repair failed plans over changing tool catalogs. We introduce Evoflux, an inference-time evolutionary search method that treats compact tool use as the repair of executable tool workflows. It evolves typed workflow graphs through structured edits, execution feedback, adaptive intensity, meta-guided redesign, and diversity pruning. On held-out MCP-Bench tasks spanning live MCP servers and 250 tools, Evoflux raises execution feasibility from roughly 3% to 17-24% across small planners. In contrast, SFT and SFT+DPO on the same search-mined data match, underperform, or collapse below zero-shot performance; ReAct reaches higher peaks, but with higher variance and token cost. These results show that execution-grounded search is more reliable under scarce teacher-trace budgets. △ Less

Submitted 10 June, 2026; originally announced June 2026.

Comments: Code is available at https://github.com/IBM/Evoflux

Deep Learning-Based Automated Quantification of TIMI Myocardial Perfusion Frame Count (DL-TMPFC) from Coronary Angiography: A Novel Framework for Rapid Assessment of Microvascular Dysfunction

Authors: Si Li, Yuanqing He, Chenkai Hu, Xiaogang Guo, Huay-Cheem Tan, Chieh Yang Koo, Xuan Zhang, Lei He, Jingyuan Zeng, Shan Xiao

Abstract: Aims: Coronary microvascular dysfunction (CMVD) affects approximately 40%-60% of patients with ischemia and non-obstructive coronary arteries, yet diagnosis remains challenging due to reliance on invasive functional testing or subjective Thrombolysis In Myocardial Infarction (TIMI) flow grade. The TIMI Myocardial Perfusion Frame Count (TMPFC) offers an objective, angiography-based quantitative mea… ▽ More Aims: Coronary microvascular dysfunction (CMVD) affects approximately 40%-60% of patients with ischemia and non-obstructive coronary arteries, yet diagnosis remains challenging due to reliance on invasive functional testing or subjective Thrombolysis In Myocardial Infarction (TIMI) flow grade. The TIMI Myocardial Perfusion Frame Count (TMPFC) offers an objective, angiography-based quantitative measure of CMVD, but its clinical translation is hindered by cumbersome manual calculation and insufficient validation. This study aims to develop and validate a deep learning-powered TMPFC calculation (DL-TMPFC), enabling integration into clinical workflows. Methods and results: DL-TMPFC framework comprised two components. A stenosis detection network first excluded obstructive coronary artery disease (CAD). A territory-aware segmentation network then identified perfusion territories and TMPFC calculation module automatically determined the first and last frames from angiographic sequences. The framework was validated in a cohort of 655 patients (445 of obstructive CAD, 100 of confirmed CMVD, 110 of control group) from three independent institutions. DL-TMPFC showed excellent agreement with expert manual measurements (bias: -0.93 frames; 95% LoA: -5.33 to +3.47; r =0.98). DL-TMPFC markedly enhanced clinical feasibility by fully automating TMPFC and removing observer dependence. Clinically, DL-TMPFC accurately identified CMVD across a full spectrum of coronary pathologies and captured the continuous severity of CMVD beyond binary classification, enabling quantitative risk stratification. Conclusion: DL-TMPFC enabled automatic, standardized, and accurate quantification of CMVD directly from routine angiography. By providing an automatic and objective measure, this tool provided immediate diagnostic information for timely recognition and management of CMVD in clinical practice. △ Less

Submitted 19 May, 2026; originally announced May 2026.

Comments: 15 pages,8 figures

MSC Class: 92C55 ACM Class: I.4.6

Structured Disagreement in Health-Literacy Annotation: Epistemic Stability, Conceptual Difficulty, and Agreement-Stratified Inference

Authors: Olga Kellert, Sriya Kondury, Candice Koo, Nemika Tyagi, Steffen Eikenberry

Abstract: Annotation pipelines in Natural Language Processing (NLP) commonly assume a single latent ground truth per instance and resolve disagreement through label aggregation. Perspectivist approaches challenge this view by treating disagreement as potentially informative rather than erroneous. We present a large-scale analysis of graded health-literacy annotations from 6,323 open-ended COVID-19 responses… ▽ More Annotation pipelines in Natural Language Processing (NLP) commonly assume a single latent ground truth per instance and resolve disagreement through label aggregation. Perspectivist approaches challenge this view by treating disagreement as potentially informative rather than erroneous. We present a large-scale analysis of graded health-literacy annotations from 6,323 open-ended COVID-19 responses collected in Ecuador and Peru. Each response was independently labeled by multiple annotators using proportional correctness scores, reflecting the degree to which responses align with normative public-health guidelines, allowing us to analyze the full distribution of judgments rather than aggregated labels. Variance decomposition shows that question-level conceptual difficulty accounts for substantially more variance than annotator identity, indicating that disagreement is structured by the task itself rather than driven by individual raters. Agreement-stratified analyses further reveal that key social-scientific effects, including country, education, and urban-rural differences, vary in magnitude and in some cases reverse direction across levels of inter-annotator agreement. These findings suggest that graded health-literacy evaluation contains both epistemically stable and unstable components, and that aggregating across them can obscure important inferential differences. We therefore argue that strong perspectivist modeling is not only conceptually justified but statistically necessary for valid inference in graded interpretive tasks. △ Less

Submitted 21 April, 2026; originally announced April 2026.

Comments: 8 pages, 5 figures

Beyond Static Benchmarks: Synthesizing Harmful Content via Persona-based Simulation for Robust Evaluation

Authors: Huije Lee, Jisu Shin, Hoyun Song, Changgeon Ko, Jong C. Park

Abstract: Static benchmarks for harmful content detection face limitations in scalability and diversity, and may also be affected by contamination from web-scale pre-training corpora. To address these issues, we propose a framework for synthesizing harmful content, leveraging persona-guided large language model (LLM) agents. Our approach constructs two-dimensional user personas by integrating demographic id… ▽ More Static benchmarks for harmful content detection face limitations in scalability and diversity, and may also be affected by contamination from web-scale pre-training corpora. To address these issues, we propose a framework for synthesizing harmful content, leveraging persona-guided large language model (LLM) agents. Our approach constructs two-dimensional user personas by integrating demographic identities and topical interests with situational harmful strategies, enabling the simulation of diverse and contextually grounded harmful interactions. We evaluate the framework along three dimensions: harmfulness, challenge level, and diversity. Both human and LLM-based evaluations confirm that our framework achieves a high harmful generation success rate. Experiments across multiple detection systems reveal that our synthetic scenarios are more challenging to detect than those in existing benchmarks. Furthermore, a multi-faceted analysis confirms that our approach achieves linguistic and topical diversity comparable to human-curated datasets, establishing our framework as an effective tool for robust stress-testing of harmful content detection systems. △ Less

Submitted 18 April, 2026; originally announced April 2026.

Comments: ACL 2026

Social Dynamics as Critical Vulnerabilities that Undermine Objective Decision-Making in LLM Collectives

Authors: Changgeon Ko, Jisu Shin, Hoyun Song, Huije Lee, Eui Jun Hwang, Jong C. Park

Abstract: Large language model (LLM) agents are increasingly acting as human delegates in multi-agent environments, where a representative agent integrates diverse peer perspectives to make a final decision. Drawing inspiration from social psychology, we investigate how the reliability of this representative agent is undermined by the social context of its network. We define four key phenomena-social confor… ▽ More Large language model (LLM) agents are increasingly acting as human delegates in multi-agent environments, where a representative agent integrates diverse peer perspectives to make a final decision. Drawing inspiration from social psychology, we investigate how the reliability of this representative agent is undermined by the social context of its network. We define four key phenomena-social conformity, perceived expertise, dominant speaker effect, and rhetorical persuasion-and systematically manipulate the number of adversaries, relative intelligence, argument length, and argumentative styles. Our experiments demonstrate that the representative agent's accuracy consistently declines as social pressure increases: larger adversarial groups, more capable peers, and longer arguments all lead to significant performance degradation. Furthermore, rhetorical strategies emphasizing credibility or logic can further sway the agent's judgment, depending on the context. These findings reveal that multi-agent systems are sensitive not only to individual reasoning but also to the social dynamics of their configuration, highlighting critical vulnerabilities in AI delegates that mirror the psychological biases observed in human group decision-making. △ Less

Submitted 2 May, 2026; v1 submitted 7 April, 2026; originally announced April 2026.

Comments: ACL 2026

From Static Templates to Dynamic Runtime Graphs: A Survey of Workflow Optimization for LLM Agents

Authors: Ling Yue, Kushal Raj Bhandari, Ching-Yun Ko, Dhaval Patel, Shuxin Lin, Nianjun Zhou, Jianxi Gao, Pin-Yu Chen, Shaowu Pan

Abstract: Large language model (LLM)-based systems are becoming increasingly popular for solving tasks by constructing executable workflows that interleave LLM calls, information retrieval, tool use, code execution, memory updates, and verification. This survey reviews recent methods for designing and optimizing such workflows, which we treat as agentic computation graphs (ACGs). We organize the literature… ▽ More Large language model (LLM)-based systems are becoming increasingly popular for solving tasks by constructing executable workflows that interleave LLM calls, information retrieval, tool use, code execution, memory updates, and verification. This survey reviews recent methods for designing and optimizing such workflows, which we treat as agentic computation graphs (ACGs). We organize the literature based on when workflow structure is determined, where structure refers to which components or agents are present, how they depend on each other, and how information flows between them. This lens distinguishes static methods, which fix a reusable workflow scaffold before deployment, from dynamic methods, which select, generate, or revise the workflow for a particular run before or during execution. We further organize prior work along three dimensions: when structure is determined, what part of the workflow is optimized, and which evaluation signals guide optimization (e.g., task metrics, verifier signals, preferences, or trace-derived feedback). We also distinguish reusable workflow templates, run-specific realized graphs, and execution traces, separating reusable design choices from the structures actually deployed in a given run and from realized runtime behavior. Finally, we outline a structure-aware evaluation perspective that complements downstream task metrics with graph-level properties, execution cost, robustness, and structural variation across inputs. Our goal is to provide a clear vocabulary, a unified framework for positioning new methods, a more comparable view of existing body of literature, and a more reproducible evaluation standard for future work in workflow optimizations for LLM agents. △ Less

Submitted 23 March, 2026; originally announced March 2026.

X-AVDT: Audio-Visual Cross-Attention for Robust Deepfake Detection

Authors: Youngseo Kim, Kwan Yun, Seokhyeon Hong, Sihun Cha, Colette Suhjung Koo, Junyong Noh

Abstract: The surge of highly realistic synthetic videos produced by contemporary generative systems has significantly increased the risk of malicious use, challenging both humans and existing detectors. Against this backdrop, we take a generator-side view and observe that internal cross-attention mechanisms in these models encode fine-grained speech-motion alignment, offering useful correspondence cues for… ▽ More The surge of highly realistic synthetic videos produced by contemporary generative systems has significantly increased the risk of malicious use, challenging both humans and existing detectors. Against this backdrop, we take a generator-side view and observe that internal cross-attention mechanisms in these models encode fine-grained speech-motion alignment, offering useful correspondence cues for forgery detection. Building on this insight, we propose X-AVDT, a robust and generalizable deepfake detector that probes generator-internal audio-visual signals accessed via DDIM inversion to expose these cues. X-AVDT extracts two complementary signals: (i) a video composite capturing inversion-induced discrepancies, and (ii) an audio-visual cross-attention feature reflecting modality alignment enforced during generation. To enable faithful cross-generator evaluation, we further introduce MMDF, a new multimodal deepfake dataset spanning diverse manipulation types and rapidly evolving synthesis paradigms, including GANs, diffusion, and flow-matching. Extensive experiments demonstrate that X-AVDT achieves leading performance on MMDF and generalizes strongly to external benchmarks and unseen generators, outperforming existing methods with accuracy improved by 13.1%. Our findings highlight the importance of leveraging internal audio-visual consistency cues for robustness to future generators in deepfake detection. △ Less

Submitted 9 March, 2026; originally announced March 2026.

Journal ref: CVPR 2026

vLLM Hook v0: A Plug-in for Programming Model Internals on vLLM

Authors: Ching-Yun Ko, Pin-Yu Chen

Abstract: Modern artificial intelligence (AI) models are deployed on inference engines to optimize runtime efficiency and resource allocation, particularly for transformer-based large language models (LLMs). The vLLM project is a major open-source library to support model serving and inference. However, the current implementation of vLLM limits programmability of the internal states of deployed models. This… ▽ More Modern artificial intelligence (AI) models are deployed on inference engines to optimize runtime efficiency and resource allocation, particularly for transformer-based large language models (LLMs). The vLLM project is a major open-source library to support model serving and inference. However, the current implementation of vLLM limits programmability of the internal states of deployed models. This prevents the use of popular test-time model alignment and enhancement methods. For example, it prevents the detection of adversarial prompts based on attention patterns or the adjustment of model responses based on activation steering. To bridge this critical gap, we present vLLM Hook, an opensource plug-in to enable the programming of internal states for vLLM models. Based on a configuration file specifying which internal states to capture, vLLM Hook provides seamless integration to vLLM and supports two essential features: passive programming and active programming. For passive programming, vLLM Hook probes the selected internal states for subsequent analysis, while keeping the model generation intact. For active programming, vLLM Hook enables efficient intervention of model generation by altering the selected internal states. In addition to presenting the core functions of vLLM Hook, in version 0, we demonstrate 3 use cases including prompt injection detection, enhanced retrieval-augmented retrieval (RAG), and activation steering. Finally, we welcome the community's contribution to improve vLLM Hook via https://github.com/ibm/vllm-hook. △ Less

Submitted 2 February, 2026; originally announced March 2026.

Learning Rate Matters: Vanilla LoRA May Suffice for LLM Fine-tuning

Authors: Yu-Ang Lee, Ching-Yun Ko, Pin-Yu Chen, Mi-Yen Yeh

Abstract: Low-Rank Adaptation (LoRA) is the prevailing approach for efficient large language model (LLM) fine-tuning. Building on this paradigm, recent studies have proposed alternative initialization strategies, architectural modifications, and optimization adjustments, reporting substantial improvements over vanilla LoRA. However, these gains are often demonstrated under fixed or narrowly tuned hyperparam… ▽ More Low-Rank Adaptation (LoRA) is the prevailing approach for efficient large language model (LLM) fine-tuning. Building on this paradigm, recent studies have proposed alternative initialization strategies, architectural modifications, and optimization adjustments, reporting substantial improvements over vanilla LoRA. However, these gains are often demonstrated under fixed or narrowly tuned hyperparameter settings, despite the known sensitivity of neural networks to training configurations. In this work, we systematically re-evaluate nine representative LoRA variants alongside vanilla LoRA through extensive hyperparameter searches over learning rate, batch size, rank, and training duration. Across tasks spanning mathematical reasoning, commonsense reasoning, code generation, and instruction following at diverse model scales, we find that different LoRA methods favor distinct learning rate ranges. Crucially, once learning rates are properly tuned, all methods achieve similar peak performance (within 1-2%), with only subtle rank-dependent behaviors. These results suggest that vanilla LoRA remains a competitive baseline and that improvements reported under a single training configuration may not reflect consistent methodological advantages. Finally, a second-order analysis attributes the differing optimal learning rate ranges to variations in the largest Hessian eigenvalue, aligning with classical learning theories. △ Less

Submitted 18 May, 2026; v1 submitted 4 February, 2026; originally announced February 2026.

Comments: Project page: https://github.com/yuang-lee/lr-matters-lora

Steering Externalities: Benign Activation Steering Unintentionally Increases Jailbreak Risk for Large Language Models

Authors: Chen Xiong, Zhiyuan He, Pin-Yu Chen, Ching-Yun Ko, Tsung-Yi Ho

Abstract: Activation steering is a practical post-training model alignment technique to enhance the utility of Large Language Models (LLMs). Prior to deploying a model as a service, developers can steer a pre-trained model toward specific behavioral objectives, such as compliance or instruction adherence, without the need for retraining. This process is as simple as adding a steering vector to the model's i… ▽ More Activation steering is a practical post-training model alignment technique to enhance the utility of Large Language Models (LLMs). Prior to deploying a model as a service, developers can steer a pre-trained model toward specific behavioral objectives, such as compliance or instruction adherence, without the need for retraining. This process is as simple as adding a steering vector to the model's internal representations. However, this capability unintentionally introduces critical and under-explored safety risks. We identify a phenomenon termed Steering Externalities, where steering vectors derived from entirely benign datasets-such as those enforcing strict compliance or specific output formats like JSON-inadvertently erode safety guardrails. Experiments reveal that these interventions act as a force multiplier, creating new vulnerabilities to jailbreaks and increasing attack success rates to over 80% on standard benchmarks by bypassing the initial safety alignment. Ultimately, our results expose a critical blind spot in deployment: benign activation steering systematically erodes the "safety margin," rendering models more vulnerable to black-box attacks and proving that inference-time utility improvements must be rigorously audited for unintended safety externalities. △ Less

Submitted 3 February, 2026; originally announced February 2026.

EARL: Entropy-Aware RL Alignment of LLMs for Reliable RTL Code Generation

Authors: Jiahe Shi, Zhengqi Gao, Ching-Yun Ko, Duane Boning

Abstract: Recent advances in large language models (LLMs) have demonstrated significant potential in hardware design automation, particularly in using natural language to synthesize Register-Transfer Level (RTL) code. Despite this progress, a gap remains between model capability and the demands of real-world RTL design, including syntax errors, functional hallucinations, and weak alignment to designer inten… ▽ More Recent advances in large language models (LLMs) have demonstrated significant potential in hardware design automation, particularly in using natural language to synthesize Register-Transfer Level (RTL) code. Despite this progress, a gap remains between model capability and the demands of real-world RTL design, including syntax errors, functional hallucinations, and weak alignment to designer intent. Reinforcement Learning with Verifiable Rewards (RLVR) offers a promising approach to bridge this gap, as hardware provides executable and formally checkable signals that can be used to further align model outputs with design intent. However, in long, structured RTL code sequences, not all tokens contribute equally to functional correctness, and naïvely spreading gradients across all tokens dilutes learning signals. A key insight from our entropy analysis in RTL generation is that only a small fraction of tokens (e.g., always, if, assign, posedge) exhibit high uncertainty and largely influence control flow and module structure. To address these challenges, we present EARL, an Entropy-Aware Reinforcement Learning framework for Verilog generation. EARL performs policy optimization using verifiable reward signals and introduces entropy-guided selective updates that gate policy gradients to high-entropy tokens. This approach preserves training stability and concentrates gradient updates on functionally important regions of code. Our experiments on VerilogEval and RTLLM show that EARL improves functional pass rates over prior LLM baselines by up to 14.7%, while reducing unnecessary updates and improving training stability. These results indicate that focusing RL on critical, high-uncertainty tokens enables more reliable and targeted policy improvement for structured RTL code generation. △ Less

Submitted 24 February, 2026; v1 submitted 15 November, 2025; originally announced November 2025.

Comments: Accepted to DAC 2026

Patching LLM Like Software: A Lightweight Method for Improving Safety Policy in Large Language Models

Authors: Huzaifa Arif, Keerthiram Murugesan, Ching-Yun Ko, Pin-Yu Chen, Payel Das, Alex Gittens

Abstract: We propose patching for large language models (LLMs) like software versions, a lightweight and modular approach for addressing safety vulnerabilities. While vendors release improved LLM versions, major releases are costly, infrequent, and difficult to tailor to customer needs, leaving released models with known safety gaps. Unlike full-model fine-tuning or major version updates, our method enables… ▽ More We propose patching for large language models (LLMs) like software versions, a lightweight and modular approach for addressing safety vulnerabilities. While vendors release improved LLM versions, major releases are costly, infrequent, and difficult to tailor to customer needs, leaving released models with known safety gaps. Unlike full-model fine-tuning or major version updates, our method enables rapid remediation by prepending a compact, learnable prefix to an existing model. This "patch" introduces only 0.003% additional parameters, yet reliably steers model behavior toward that of a safer reference model. Across three critical domains (toxicity mitigation, bias reduction, and harmfulness refusal) policy patches achieve safety improvements comparable to next-generation safety-aligned models while preserving fluency. Our results demonstrate that LLMs can be "patched" much like software, offering vendors and practitioners a practical mechanism for distributing scalable, efficient, and composable safety updates between major model releases. △ Less

Submitted 27 April, 2026; v1 submitted 11 November, 2025; originally announced November 2025.

Language over Content: Tracing Cultural Understanding in Multilingual Large Language Models

Authors: Seungho Cho, Changgeon Ko, Eui Jun Hwang, Junmyeong Lee, Huije Lee, Jong C. Park

Abstract: Large language models (LLMs) are increasingly used across diverse cultural contexts, making accurate cultural understanding essential. Prior evaluations have mostly focused on output-level performance, obscuring the factors that drive differences in responses, while studies using circuit analysis have covered few languages and rarely focused on culture. In this work, we trace LLMs' internal cultur… ▽ More Large language models (LLMs) are increasingly used across diverse cultural contexts, making accurate cultural understanding essential. Prior evaluations have mostly focused on output-level performance, obscuring the factors that drive differences in responses, while studies using circuit analysis have covered few languages and rarely focused on culture. In this work, we trace LLMs' internal cultural understanding mechanisms by measuring activation path overlaps when answering semantically equivalent questions under two conditions: varying the target country while fixing the question language, and varying the question language while fixing the country. We also use same-language country pairs to disentangle language from cultural aspects. Results show that internal paths overlap more for same-language, cross-country questions than for cross-language, same-country questions, indicating strong language-specific patterns. Notably, the South Korea-North Korea pair exhibits low overlap and high variability, showing that linguistic similarity does not guarantee aligned internal representation. △ Less

Submitted 10 November, 2025; v1 submitted 18 October, 2025; originally announced October 2025.

Comments: Accepted to CIKM 2025 Workshop on Human Centric AI

RoleConflictBench: A Benchmark of Role Conflict Scenarios for Evaluating LLMs' Contextual Sensitivity

Authors: Jisu Shin, Hoyun Song, Juhyun Oh, Changgeon Ko, Eunsu Kim, Chani Jung, Alice Oh

Abstract: People often encounter role conflicts -- social dilemmas where the expectations of multiple roles clash and cannot be simultaneously fulfilled. As large language models (LLMs) increasingly navigate these social dynamics, a critical research question emerges. When faced with such dilemmas, do LLMs prioritize dynamic contextual cues or the learned preferences? To address this, we introduce RoleConfl… ▽ More People often encounter role conflicts -- social dilemmas where the expectations of multiple roles clash and cannot be simultaneously fulfilled. As large language models (LLMs) increasingly navigate these social dynamics, a critical research question emerges. When faced with such dilemmas, do LLMs prioritize dynamic contextual cues or the learned preferences? To address this, we introduce RoleConflictBench, a novel benchmark designed to measure the contextual sensitivity of LLMs in role conflict scenarios. To enable objective evaluation within this subjective domain, we employ situational urgency as a constraint for decision-making. We construct the dataset through a three-stage pipeline that generates over 13,000 realistic scenarios across 65 roles in five social domains by systematically varying the urgency of competing situations. This controlled setup enables us to quantitatively measure contextual sensitivity, determining whether model decisions align with the situational contexts or are overridden by the learned role preferences. Our analysis of 10 LLMs reveals that models substantially deviate from this objective baseline. Instead of responding to dynamic contextual cues, their decisions are predominantly governed by the preferences toward specific social roles. △ Less

Submitted 17 April, 2026; v1 submitted 30 September, 2025; originally announced September 2025.

Comments: Accepted to Findings of ACL 2026

STRIDE: Subset-Free Functional Decomposition for XAI in Tabular Settings

Authors: Chaeyun Ko

Abstract: Most explainable AI (XAI) frameworks are limited in their expressiveness, summarizing complex feature effects as single scalar values φ_i. This approach answers "what" features are important but fails to reveal "how" they interact. Furthermore, methods that attempt to capture interactions, like those based on Shapley values, often face an exponential computational cost. We present STRIDE, a scalab… ▽ More Most explainable AI (XAI) frameworks are limited in their expressiveness, summarizing complex feature effects as single scalar values φ_i. This approach answers "what" features are important but fails to reveal "how" they interact. Furthermore, methods that attempt to capture interactions, like those based on Shapley values, often face an exponential computational cost. We present STRIDE, a scalable framework that addresses both limitations by reframing explanation as a subset-enumeration-free, orthogonal "functional decomposition" in a Reproducing Kernel Hilbert Space (RKHS). In the tabular setups we study, STRIDE analytically computes functional components f_S(x_S) via a recursive kernel-centering procedure. The approach is model-agnostic and theoretically grounded with results on orthogonality and L^2 convergence. In tabular benchmarks (10 datasets, median over 10 seeds), STRIDE attains a 3.0 times median speedup over TreeSHAP and a mean R^2=0.93 for reconstruction. We also introduce "component surgery", a diagnostic that isolates a learned interaction and quantifies its contribution; on California Housing, removing a single interaction reduces test R^2 from 0.019 to 0.027. △ Less

Submitted 14 September, 2025; v1 submitted 10 September, 2025; originally announced September 2025.

Comments: Major revision for submission to ICLR 2026. Substantially revised abstract, introduction, and discussion. Added new 'component surgery' analysis and updated benchmark results for clarity. (12 pages, 2 figures)

Biased Local SGD for Efficient Deep Learning on Heterogeneous Systems

Authors: Jihyun Lim, Junhyuk Jo, Chanhyeok Ko, Young Min Go, Jimin Hwa, Sunwoo Lee

Abstract: Most parallel neural network training methods assume homogeneous computing resources. For example, synchronous data-parallel SGD suffers from significant synchronization overhead under heterogeneous workloads, often forcing practitioners to rely only on the fastest devices (e.g., GPUs). In this work, we study local SGD for efficient parallel training on heterogeneous systems. We show that intentio… ▽ More Most parallel neural network training methods assume homogeneous computing resources. For example, synchronous data-parallel SGD suffers from significant synchronization overhead under heterogeneous workloads, often forcing practitioners to rely only on the fastest devices (e.g., GPUs). In this work, we study local SGD for efficient parallel training on heterogeneous systems. We show that intentionally introducing bias in data sampling and model aggregation can effectively harmonize slower CPUs with faster GPUs. Our extensive empirical results demonstrate that a carefully controlled bias significantly accelerates local SGD while achieving comparable or even higher accuracy than synchronous SGD under the same epoch budget. For instance, our method trains ResNet20 on CIFAR-10 with 2 CPUs and 8 GPUs up to 32x faster than synchronous SGD, with nearly identical accuracy. These results provide practical insights into how to flexibly utilize diverse compute resources for deep learning. △ Less

Submitted 22 February, 2026; v1 submitted 11 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

Does Rationale Quality Matter? Enhancing Mental Disorder Detection via Selective Reasoning Distillation

Authors: Hoyun Song, Huije Lee, Jisu Shin, Sukmin Cho, Changgeon Ko, Jong C. Park

Abstract: The detection of mental health problems from social media and the interpretation of these results have been extensively explored. Research has shown that incorporating clinical symptom information into a model enhances domain expertise, improving its detection and interpretation performance. While large language models (LLMs) are shown to be effective for generating explanatory rationales in menta… ▽ More The detection of mental health problems from social media and the interpretation of these results have been extensively explored. Research has shown that incorporating clinical symptom information into a model enhances domain expertise, improving its detection and interpretation performance. While large language models (LLMs) are shown to be effective for generating explanatory rationales in mental health detection, their substantially large parameter size and high computational cost limit their practicality. Reasoning distillation transfers this ability to smaller language models (SLMs), but inconsistencies in the relevance and domain alignment of LLM-generated rationales pose a challenge. This paper investigates how rationale quality impacts SLM performance in mental health detection and explanation generation. We hypothesize that ensuring high-quality and domain-relevant rationales enhances the distillation. To this end, we propose a framework that selects rationales based on their alignment with expert clinical reasoning. Experiments show that our quality-focused approach significantly enhances SLM performance in both mental disorder detection and rationale generation. This work highlights the importance of rationale quality and offers an insightful framework for knowledge transfer in mental health applications. △ Less

Submitted 26 May, 2025; originally announced May 2025.

A Multi-Task Benchmark for Abusive Language Detection in Low-Resource Settings

Authors: Fitsum Gaim, Hoyun Song, Huije Lee, Changgeon Ko, Eui Jun Hwang, Jong C. Park

Abstract: Content moderation research has recently made significant advances, but remains limited in serving the majority of the world's languages due to the lack of resources, leaving millions of vulnerable users to online hostility. This work presents a large-scale human-annotated multi-task benchmark dataset for abusive language detection in Tigrinya social media with joint annotations for three tasks: a… ▽ More Content moderation research has recently made significant advances, but remains limited in serving the majority of the world's languages due to the lack of resources, leaving millions of vulnerable users to online hostility. This work presents a large-scale human-annotated multi-task benchmark dataset for abusive language detection in Tigrinya social media with joint annotations for three tasks: abusiveness, sentiment, and topic classification. The dataset comprises 13,717 YouTube comments annotated by nine native speakers, collected from 7,373 videos with a total of over 1.2 billion views across 51 channels. We developed an iterative term clustering approach for effective data selection. Recognizing that around 64% of Tigrinya social media content uses Romanized transliterations rather than native Ge'ez script, our dataset accommodates both writing systems to reflect actual language use. We establish strong baselines across the tasks in the benchmark, while leaving significant challenges for future contributions. Our experiments demonstrate that small fine-tuned models outperform prompted frontier large language models (LLMs) in the low-resource setting, achieving 86.67% F1 in abusiveness detection (7+ points over best LLM), and maintain stronger performance in all other tasks. The benchmark is made public to promote research on online safety. △ Less

Submitted 25 October, 2025; v1 submitted 17 May, 2025; originally announced May 2025.

Comments: Accepted at NeurIPS 2025

ACM Class: I.2.7

Can Memory-Augmented Language Models Generalize on Reasoning-in-a-Haystack Tasks?

Authors: Payel Das, Ching-Yun Ko, Sihui Dai, Georgios Kollias, Subhajit Chaudhury, Aurelie Lozano

Abstract: Large language models often expose their brittleness in reasoning tasks, especially while executing long chains of reasoning over context. We propose MemReasoner, a new and simple memory-augmented LLM architecture, in which the memory learns the relative order of facts in context, and enables hopping over them, while the decoder selectively attends to the memory. MemReasoner is trained end-to-end,… ▽ More Large language models often expose their brittleness in reasoning tasks, especially while executing long chains of reasoning over context. We propose MemReasoner, a new and simple memory-augmented LLM architecture, in which the memory learns the relative order of facts in context, and enables hopping over them, while the decoder selectively attends to the memory. MemReasoner is trained end-to-end, with optional supporting fact supervision of varying degrees. We train MemReasoner, along with existing memory-augmented transformer models and a state-space model, on two distinct synthetic multi-hop reasoning tasks. Experiments performed under a variety of challenging scenarios, including the presence of long distractor text or target answer changes in test set, show strong generalization of MemReasoner on both single- and two-hop tasks. This generalization of MemReasoner is achieved using none-to-weak supporting fact supervision (using none and 1\% of supporting facts for one- and two-hop tasks, respectively). In contrast, baseline models overall struggle to generalize and benefit far less from using full supporting fact supervision. The results highlight the importance of explicit memory mechanisms, combined with additional weak supervision, for improving large language model's context processing ability toward reasoning tasks. △ Less

Submitted 10 March, 2025; originally announced March 2025.

STAR: Spectral Truncation and Rescale for Model Merging

Authors: Yu-Ang Lee, Ching-Yun Ko, Tejaswini Pedapati, I-Hsin Chung, Mi-Yen Yeh, Pin-Yu Chen

Abstract: Model merging is an efficient way of obtaining a multi-task model from several pretrained models without further fine-tuning, and it has gained attention in various domains, including natural language processing (NLP). Despite the efficiency, a key challenge in model merging is the seemingly inevitable decrease in task performance as the number of models increases. In this paper, we propose… ▽ More Model merging is an efficient way of obtaining a multi-task model from several pretrained models without further fine-tuning, and it has gained attention in various domains, including natural language processing (NLP). Despite the efficiency, a key challenge in model merging is the seemingly inevitable decrease in task performance as the number of models increases. In this paper, we propose $\mathbf{S}$pectral $\mathbf{T}$runcation $\mathbf{A}$nd $\mathbf{R}$escale (STAR) that aims at mitigating ``merging conflicts'' by truncating small components in the respective spectral spaces, which is followed by an automatic parameter rescaling scheme to retain the nuclear norm of the original matrix. STAR requires no additional inference on original training data and is robust to hyperparamater choice. We demonstrate the effectiveness of STAR through extensive model merging cases on diverse NLP tasks. Specifically, STAR works robustly across varying model sizes, and can outperform baselines by 4.2$\%$ when merging 12 models on Flan-T5. Our code is publicly available at https://github.com/IBM/STAR. △ Less

Submitted 14 February, 2025; originally announced February 2025.

Comments: Accepted to NAACL 2025

Artificial Intelligence to Assess Dental Findings from Panoramic Radiographs -- A Multinational Study

Authors: Yin-Chih Chelsea Wang, Tsao-Lun Chen, Shankeeth Vinayahalingam, Tai-Hsien Wu, Chu Wei Chang, Hsuan Hao Chang, Hung-Jen Wei, Mu-Hsiung Chen, Ching-Chang Ko, David Anssari Moin, Bram van Ginneken, Tong Xi, Hsiao-Cheng Tsai, Min-Huey Chen, Tzu-Ming Harry Hsu, Hye Chou

Abstract: Dental panoramic radiographs (DPRs) are widely used in clinical practice for comprehensive oral assessment but present challenges due to overlapping structures and time constraints in interpretation. This study aimed to establish a solid baseline for the AI-automated assessment of findings in DPRs by developing, evaluating an AI system, and comparing its performance with that of human readers ac… ▽ More Dental panoramic radiographs (DPRs) are widely used in clinical practice for comprehensive oral assessment but present challenges due to overlapping structures and time constraints in interpretation. This study aimed to establish a solid baseline for the AI-automated assessment of findings in DPRs by developing, evaluating an AI system, and comparing its performance with that of human readers across multinational data sets. We analyzed 6,669 DPRs from three data sets (the Netherlands, Brazil, and Taiwan), focusing on 8 types of dental findings. The AI system combined object detection and semantic segmentation techniques for per-tooth finding identification. Performance metrics included sensitivity, specificity, and area under the receiver operating characteristic curve (AUC-ROC). AI generalizability was tested across data sets, and performance was compared with human dental practitioners. The AI system demonstrated comparable or superior performance to human readers, particularly +67.9% (95% CI: 54.0%-81.9%; p < .001) sensitivity for identifying periapical radiolucencies and +4.7% (95% CI: 1.4%-8.0%; p = .008) sensitivity for identifying missing teeth. The AI achieved a macro-averaged AUC-ROC of 96.2% (95% CI: 94.6%-97.8%) across 8 findings. AI agreements with the reference were comparable to inter-human agreements in 7 of 8 findings except for caries (p = .024). The AI system demonstrated robust generalization across diverse imaging and demographic settings and processed images 79 times faster (95% CI: 75-82) than human readers. The AI system effectively assessed findings in DPRs, achieving performance on par with or better than human experts while significantly reducing interpretation time. These results highlight the potential for integrating AI into clinical workflows to improve diagnostic efficiency and accuracy, and patient management. △ Less

Submitted 14 February, 2025; originally announced February 2025.

Understanding and Improving Training-Free AI-Generated Image Detections with Vision Foundation Models

Authors: Chung-Ting Tsai, Ching-Yun Ko, I-Hsin Chung, Yu-Chiang Frank Wang, Pin-Yu Chen

Abstract: The rapid advancement of generative models has introduced serious risks, including deepfake techniques for facial synthesis and editing. Traditional approaches rely on training classifiers and enhancing generalizability through various feature extraction techniques. Meanwhile, training-free detection methods address issues like limited data and overfitting by directly leveraging statistical proper… ▽ More The rapid advancement of generative models has introduced serious risks, including deepfake techniques for facial synthesis and editing. Traditional approaches rely on training classifiers and enhancing generalizability through various feature extraction techniques. Meanwhile, training-free detection methods address issues like limited data and overfitting by directly leveraging statistical properties from vision foundation models to distinguish between real and fake images. The current leading training-free approach, RIGID, utilizes DINOv2 sensitivity to perturbations in image space for detecting fake images, with fake image embeddings exhibiting greater sensitivity than those of real images. This observation prompts us to investigate how detection performance varies across model backbones, perturbation types, and datasets. Our experiments reveal that detection performance is closely linked to model robustness, with self-supervised (SSL) models providing more reliable representations. While Gaussian noise effectively detects general objects, it performs worse on facial images, whereas Gaussian blur is more effective due to potential frequency artifacts. To further improve detection, we introduce Contrastive Blur, which enhances performance on facial images, and MINDER (MINimum distance DetEctoR), which addresses noise type bias, balancing performance across domains. Beyond performance gains, our work offers valuable insights for both the generative and detection communities, contributing to a deeper understanding of model robustness property utilized for deepfake detection. △ Less

Submitted 28 November, 2024; originally announced November 2024.

Different Bias Under Different Criteria: Assessing Bias in LLMs with a Fact-Based Approach

Authors: Changgeon Ko, Jisu Shin, Hoyun Song, Jeongyeon Seo, Jong C. Park

Abstract: Large language models (LLMs) often reflect real-world biases, leading to efforts to mitigate these effects and make the models unbiased. Achieving this goal requires defining clear criteria for an unbiased state, with any deviation from these criteria considered biased. Some studies define an unbiased state as equal treatment across diverse demographic groups, aiming for balanced outputs from LLMs… ▽ More Large language models (LLMs) often reflect real-world biases, leading to efforts to mitigate these effects and make the models unbiased. Achieving this goal requires defining clear criteria for an unbiased state, with any deviation from these criteria considered biased. Some studies define an unbiased state as equal treatment across diverse demographic groups, aiming for balanced outputs from LLMs. However, differing perspectives on equality and the importance of pluralism make it challenging to establish a universal standard. Alternatively, other approaches propose using fact-based criteria for more consistent and objective evaluations, though these methods have not yet been fully applied to LLM bias assessments. Thus, there is a need for a metric with objective criteria that offers a distinct perspective from equality-based approaches. Motivated by this need, we introduce a novel metric to assess bias using fact-based criteria and real-world statistics. In this paper, we conducted a human survey demonstrating that humans tend to perceive LLM outputs more positively when they align closely with real-world demographic distributions. Evaluating various LLMs with our proposed metric reveals that model bias varies depending on the criteria used, highlighting the need for multi-perspective assessment. △ Less

Submitted 26 November, 2024; originally announced November 2024.

Comments: Accepted in NeurIPS 2024 Workshop on Socially Responsible Language Modelling Research (SoLaR)

Attention Tracker: Detecting Prompt Injection Attacks in LLMs

Authors: Kuo-Han Hung, Ching-Yun Ko, Ambrish Rawat, I-Hsin Chung, Winston H. Hsu, Pin-Yu Chen

Abstract: Large Language Models (LLMs) have revolutionized various domains but remain vulnerable to prompt injection attacks, where malicious inputs manipulate the model into ignoring original instructions and executing designated action. In this paper, we investigate the underlying mechanisms of these attacks by analyzing the attention patterns within LLMs. We introduce the concept of the distraction effec… ▽ More Large Language Models (LLMs) have revolutionized various domains but remain vulnerable to prompt injection attacks, where malicious inputs manipulate the model into ignoring original instructions and executing designated action. In this paper, we investigate the underlying mechanisms of these attacks by analyzing the attention patterns within LLMs. We introduce the concept of the distraction effect, where specific attention heads, termed important heads, shift focus from the original instruction to the injected instruction. Building on this discovery, we propose Attention Tracker, a training-free detection method that tracks attention patterns on instruction to detect prompt injection attacks without the need for additional LLM inference. Our method generalizes effectively across diverse models, datasets, and attack types, showing an AUROC improvement of up to 10.0% over existing methods, and performs well even on small LLMs. We demonstrate the robustness of our approach through extensive evaluations and provide insights into safeguarding LLM-integrated systems from prompt injection vulnerabilities. △ Less

Submitted 22 April, 2025; v1 submitted 1 November, 2024; originally announced November 2024.

Comments: Project page: https://huggingface.co/spaces/TrustSafeAI/Attention-Tracker

Medical Imaging Complexity and its Effects on GAN Performance

Authors: William Cagas, Chan Ko, Blake Hsiao, Shryuk Grandhi, Rishi Bhattacharya, Kevin Zhu, Michael Lam

Abstract: The proliferation of machine learning models in diverse clinical applications has led to a growing need for high-fidelity, medical image training data. Such data is often scarce due to cost constraints and privacy concerns. Alleviating this burden, medical image synthesis via generative adversarial networks (GANs) emerged as a powerful method for synthetically generating photo-realistic images bas… ▽ More The proliferation of machine learning models in diverse clinical applications has led to a growing need for high-fidelity, medical image training data. Such data is often scarce due to cost constraints and privacy concerns. Alleviating this burden, medical image synthesis via generative adversarial networks (GANs) emerged as a powerful method for synthetically generating photo-realistic images based on existing sets of real medical images. However, the exact image set size required to efficiently train such a GAN is unclear. In this work, we experimentally establish benchmarks that measure the relationship between a sample dataset size and the fidelity of the generated images, given the dataset's distribution of image complexities. We analyze statistical metrics based on delentropy, an image complexity measure rooted in Shannon's entropy in information theory. For our pipeline, we conduct experiments with two state-of-the-art GANs, StyleGAN 3 and SPADE-GAN, trained on multiple medical imaging datasets with variable sample sizes. Across both GANs, general performance improved with increasing training set size but suffered with increasing complexity. △ Less

Submitted 23 October, 2024; originally announced October 2024.

Comments: Accepted to ACCV, Workshop on Generative AI for Synthetic Medical Data

Large Language Models can be Strong Self-Detoxifiers

Authors: Ching-Yun Ko, Pin-Yu Chen, Payel Das, Youssef Mroueh, Soham Dan, Georgios Kollias, Subhajit Chaudhury, Tejaswini Pedapati, Luca Daniel

Abstract: Reducing the likelihood of generating harmful and toxic output is an essential task when aligning large language models (LLMs). Existing methods mainly rely on training an external reward model (i.e., another language model) or fine-tuning the LLM using self-generated data to influence the outcome. In this paper, we show that LLMs have the capability of self-detoxification without the use of an ad… ▽ More Reducing the likelihood of generating harmful and toxic output is an essential task when aligning large language models (LLMs). Existing methods mainly rely on training an external reward model (i.e., another language model) or fine-tuning the LLM using self-generated data to influence the outcome. In this paper, we show that LLMs have the capability of self-detoxification without the use of an additional reward model or re-training. We propose \textit{Self-disciplined Autoregressive Sampling (SASA)}, a lightweight controlled decoding algorithm for toxicity reduction of LLMs. SASA leverages the contextual representations from an LLM to learn linear subspaces characterizing toxic v.s. non-toxic output in analytical forms. When auto-completing a response token-by-token, SASA dynamically tracks the margin of the current output to steer the generation away from the toxic subspace, by adjusting the autoregressive sampling strategy. Evaluated on LLMs of different scale and nature, namely Llama-3.1-Instruct (8B), Llama-2 (7B), and GPT2-L models with the RealToxicityPrompts, BOLD, and AttaQ benchmarks, SASA markedly enhances the quality of the generated sentences relative to the original models and attains comparable performance to state-of-the-art detoxification techniques, significantly reducing the toxicity level by only using the LLM's internal representations. △ Less

Submitted 4 October, 2024; originally announced October 2024.

Comments: 20 pages

EQ-CBM: A Probabilistic Concept Bottleneck with Energy-based Models and Quantized Vectors

Authors: Sangwon Kim, Dasom Ahn, Byoung Chul Ko, In-su Jang, Kwang-Ju Kim

Abstract: The demand for reliable AI systems has intensified the need for interpretable deep neural networks. Concept bottleneck models (CBMs) have gained attention as an effective approach by leveraging human-understandable concepts to enhance interpretability. However, existing CBMs face challenges due to deterministic concept encoding and reliance on inconsistent concepts, leading to inaccuracies. We pro… ▽ More The demand for reliable AI systems has intensified the need for interpretable deep neural networks. Concept bottleneck models (CBMs) have gained attention as an effective approach by leveraging human-understandable concepts to enhance interpretability. However, existing CBMs face challenges due to deterministic concept encoding and reliance on inconsistent concepts, leading to inaccuracies. We propose EQ-CBM, a novel framework that enhances CBMs through probabilistic concept encoding using energy-based models (EBMs) with quantized concept activation vectors (qCAVs). EQ-CBM effectively captures uncertainties, thereby improving prediction reliability and accuracy. By employing qCAVs, our method selects homogeneous vectors during concept encoding, enabling more decisive task performance and facilitating higher levels of human intervention. Empirical results using benchmark datasets demonstrate that our approach outperforms the state-of-the-art in both concept and task accuracy. △ Less

Submitted 22 September, 2024; originally announced September 2024.

Comments: Accepted by ACCV 2024

SignBLEU: Automatic Evaluation of Multi-channel Sign Language Translation

Authors: Jung-Ho Kim, Mathew Huerta-Enochian, Changyong Ko, Du Hui Lee

Abstract: Sign languages are multi-channel languages that communicate information through not just the hands (manual signals) but also facial expressions and upper body movements (non-manual signals). However, since automatic sign language translation is usually performed by generating a single sequence of glosses, researchers eschew non-manual and co-occurring manual signals in favor of a simplified list o… ▽ More Sign languages are multi-channel languages that communicate information through not just the hands (manual signals) but also facial expressions and upper body movements (non-manual signals). However, since automatic sign language translation is usually performed by generating a single sequence of glosses, researchers eschew non-manual and co-occurring manual signals in favor of a simplified list of manual glosses. This can lead to significant information loss and ambiguity. In this paper, we introduce a new task named multi-channel sign language translation (MCSLT) and present a novel metric, SignBLEU, designed to capture multiple signal channels. We validated SignBLEU on a system-level task using three sign language corpora with varied linguistic structures and transcription methodologies and examined its correlation with human judgment through two segment-level tasks. We found that SignBLEU consistently correlates better with human judgment than competing metrics. To facilitate further MCSLT research, we report benchmark scores for the three sign language corpora and release the source code for SignBLEU at https://github.com/eq4all-projects/SignBLEU. △ Less

Submitted 10 June, 2024; originally announced June 2024.

Comments: Published in LREC-Coling 2024

Scene Graph Generation Strategy with Co-occurrence Knowledge and Learnable Term Frequency

Authors: Hyeongjin Kim, Sangwon Kim, Dasom Ahn, Jong Taek Lee, Byoung Chul Ko

Abstract: Scene graph generation (SGG) is an important task in image understanding because it represents the relationships between objects in an image as a graph structure, making it possible to understand the semantic relationships between objects intuitively. Previous SGG studies used a message-passing neural networks (MPNN) to update features, which can effectively reflect information about surrounding o… ▽ More Scene graph generation (SGG) is an important task in image understanding because it represents the relationships between objects in an image as a graph structure, making it possible to understand the semantic relationships between objects intuitively. Previous SGG studies used a message-passing neural networks (MPNN) to update features, which can effectively reflect information about surrounding objects. However, these studies have failed to reflect the co-occurrence of objects during SGG generation. In addition, they only addressed the long-tail problem of the training dataset from the perspectives of sampling and learning methods. To address these two problems, we propose CooK, which reflects the Co-occurrence Knowledge between objects, and the learnable term frequency-inverse document frequency (TF-l-IDF) to solve the long-tail problem. We applied the proposed model to the SGG benchmark dataset, and the results showed a performance improvement of up to 3.8% compared with existing state-of-the-art models in SGGen subtask. The proposed method exhibits generalization ability from the results obtained, showing uniform performance improvement for all MPNN models. △ Less

Submitted 21 May, 2024; originally announced May 2024.

Comments: Accepted by ICML2024

Semantic Scene Graph Generation Based on an Edge Dual Scene Graph and Message Passing Neural Network

Authors: Hyeongjin Kim, Sangwon Kim, Jong Taek Lee, Byoung Chul Ko

Abstract: Along with generative AI, interest in scene graph generation (SGG), which comprehensively captures the relationships and interactions between objects in an image and creates a structured graph-based representation, has significantly increased in recent years. However, relying on object-centric and dichotomous relationships, existing SGG methods have a limited ability to accurately predict detailed… ▽ More Along with generative AI, interest in scene graph generation (SGG), which comprehensively captures the relationships and interactions between objects in an image and creates a structured graph-based representation, has significantly increased in recent years. However, relying on object-centric and dichotomous relationships, existing SGG methods have a limited ability to accurately predict detailed relationships. To solve these problems, a new approach to the modeling multiobject relationships, called edge dual scene graph generation (EdgeSGG), is proposed herein. EdgeSGG is based on a edge dual scene graph and Dual Message Passing Neural Network (DualMPNN), which can capture rich contextual interactions between unconstrained objects. To facilitate the learning of edge dual scene graphs with a symmetric graph structure, the proposed DualMPNN learns both object- and relation-centric features for more accurately predicting relation-aware contexts and allows fine-grained relational updates between objects. A comparative experiment with state-of-the-art (SoTA) methods was conducted using two public datasets for SGG operations and six metrics for three subtasks. Compared with SoTA approaches, the proposed model exhibited substantial performance improvements across all SGG subtasks. Furthermore, experiment on long-tail distributions revealed that incorporating the relationships between objects effectively mitigates existing long-tail problems. △ Less

Submitted 2 November, 2023; originally announced November 2023.

Sample-Specific Debiasing for Better Image-Text Models

Authors: Peiqi Wang, Yingcheng Liu, Ching-Yun Ko, William M. Wells, Seth Berkowitz, Steven Horng, Polina Golland

Abstract: Self-supervised representation learning on image-text data facilitates crucial medical applications, such as image classification, visual grounding, and cross-modal retrieval. One common approach involves contrasting semantically similar (positive) and dissimilar (negative) pairs of data points. Drawing negative samples uniformly from the training data set introduces false negatives, i.e., samples… ▽ More Self-supervised representation learning on image-text data facilitates crucial medical applications, such as image classification, visual grounding, and cross-modal retrieval. One common approach involves contrasting semantically similar (positive) and dissimilar (negative) pairs of data points. Drawing negative samples uniformly from the training data set introduces false negatives, i.e., samples that are treated as dissimilar but belong to the same class. In healthcare data, the underlying class distribution is nonuniform, implying that false negatives occur at a highly variable rate. To improve the quality of learned representations, we develop a novel approach that corrects for false negatives. Our method can be viewed as a variant of debiased contrastive learning that uses estimated sample-specific class probabilities. We provide theoretical analysis of the objective function and demonstrate the proposed approach on both image and paired image-text data sets. Our experiments illustrate empirical advantages of sample-specific debiasing. △ Less

Submitted 12 August, 2023; v1 submitted 25 April, 2023; originally announced April 2023.

Comments: Machine Learning for Healthcare Conference 2023

Cross-Modal Learning with 3D Deformable Attention for Action Recognition

Authors: Sangwon Kim, Dasom Ahn, Byoung Chul Ko

Abstract: An important challenge in vision-based action recognition is the embedding of spatiotemporal features with two or more heterogeneous modalities into a single feature. In this study, we propose a new 3D deformable transformer for action recognition with adaptive spatiotemporal receptive fields and a cross-modal learning scheme. The 3D deformable transformer consists of three attention modules: 3D d… ▽ More An important challenge in vision-based action recognition is the embedding of spatiotemporal features with two or more heterogeneous modalities into a single feature. In this study, we propose a new 3D deformable transformer for action recognition with adaptive spatiotemporal receptive fields and a cross-modal learning scheme. The 3D deformable transformer consists of three attention modules: 3D deformability, local joint stride, and temporal stride attention. The two cross-modal tokens are input into the 3D deformable attention module to create a cross-attention token with a reflected spatiotemporal correlation. Local joint stride attention is applied to spatially combine attention and pose tokens. Temporal stride attention temporally reduces the number of input tokens in the attention module and supports temporal expression learning without the simultaneous use of all tokens. The deformable transformer iterates L-times and combines the last cross-modal token for classification. The proposed 3D deformable transformer was tested on the NTU60, NTU120, FineGYM, and PennAction datasets, and showed results better than or similar to pre-trained state-of-the-art methods even without a pre-training process. In addition, by visualizing important joints and correlations during action recognition through spatial joint and temporal stride attention, the possibility of achieving an explainable potential for action recognition is presented. △ Less

Submitted 17 August, 2023; v1 submitted 11 December, 2022; originally announced December 2022.

Comments: Accepted by ICCV2023

STAR-Transformer: A Spatio-temporal Cross Attention Transformer for Human Action Recognition

Authors: Dasom Ahn, Sangwon Kim, Hyunsu Hong, Byoung Chul Ko

Abstract: In action recognition, although the combination of spatio-temporal videos and skeleton features can improve the recognition performance, a separate model and balancing feature representation for cross-modal data are required. To solve these problems, we propose Spatio-TemporAl cRoss (STAR)-transformer, which can effectively represent two cross-modal features as a recognizable vector. First, from t… ▽ More In action recognition, although the combination of spatio-temporal videos and skeleton features can improve the recognition performance, a separate model and balancing feature representation for cross-modal data are required. To solve these problems, we propose Spatio-TemporAl cRoss (STAR)-transformer, which can effectively represent two cross-modal features as a recognizable vector. First, from the input video and skeleton sequence, video frames are output as global grid tokens and skeletons are output as joint map tokens, respectively. These tokens are then aggregated into multi-class tokens and input into STAR-transformer. The STAR-transformer encoder layer consists of a full self-attention (FAttn) module and a proposed zigzag spatio-temporal attention (ZAttn) module. Similarly, the continuous decoder consists of a FAttn module and a proposed binary spatio-temporal attention (BAttn) module. STAR-transformer learns an efficient multi-feature representation of the spatio-temporal features by properly arranging pairings of the FAttn, ZAttn, and BAttn modules. Experimental results on the Penn-Action, NTU RGB+D 60, and 120 datasets show that the proposed method achieves a promising improvement in performance in comparison to previous state-of-the-art methods. △ Less

Submitted 14 October, 2022; originally announced October 2022.

Comments: Accepted by WACV 2023

MSC Class: 68T07

Demand Layering for Real-Time DNN Inference with Minimized Memory Usage

Authors: Mingoo Ji, Saehanseul Yi, Changjin Koo, Sol Ahn, Dongjoo Seo, Nikil Dutt, Jong-Chan Kim

Abstract: When executing a deep neural network (DNN), its model parameters are loaded into GPU memory before execution, incurring a significant GPU memory burden. There are studies that reduce GPU memory usage by exploiting CPU memory as a swap device. However, this approach is not applicable in most embedded systems with integrated GPUs where CPU and GPU share a common memory. In this regard, we present De… ▽ More When executing a deep neural network (DNN), its model parameters are loaded into GPU memory before execution, incurring a significant GPU memory burden. There are studies that reduce GPU memory usage by exploiting CPU memory as a swap device. However, this approach is not applicable in most embedded systems with integrated GPUs where CPU and GPU share a common memory. In this regard, we present Demand Layering, which employs a fast solid-state drive (SSD) as a co-running partner of a GPU and exploits the layer-by-layer execution of DNNs. In our approach, a DNN is loaded and executed in a layer-by-layer manner, minimizing the memory usage to the order of a single layer. Also, we developed a pipeline architecture that hides most additional delays caused by the interleaved parameter loadings alongside layer executions. Our implementation shows a 96.5% memory reduction with just 14.8% delay overhead on average for representative DNNs. Furthermore, by exploiting the memory-delay tradeoff, near-zero delay overhead (under 1 ms) can be achieved with a slightly increased memory usage (still an 88.4% reduction), showing the great potential of Demand Layering. △ Less

Submitted 8 October, 2022; originally announced October 2022.

Comments: 14 pages, 16 figures. Accepted to the 43rd IEEE Real-Time Systems Symposium (RTSS), 2022

SynBench: Task-Agnostic Benchmarking of Pretrained Representations using Synthetic Data

Authors: Ching-Yun Ko, Pin-Yu Chen, Jeet Mohapatra, Payel Das, Luca Daniel

Abstract: Recent success in fine-tuning large models, that are pretrained on broad data at scale, on downstream tasks has led to a significant paradigm shift in deep learning, from task-centric model design to task-agnostic representation learning and task-specific fine-tuning. As the representations of pretrained models are used as a foundation for different downstream tasks, this paper proposes a new task… ▽ More Recent success in fine-tuning large models, that are pretrained on broad data at scale, on downstream tasks has led to a significant paradigm shift in deep learning, from task-centric model design to task-agnostic representation learning and task-specific fine-tuning. As the representations of pretrained models are used as a foundation for different downstream tasks, this paper proposes a new task-agnostic framework, \textit{SynBench}, to measure the quality of pretrained representations using synthetic data. We set up a reference by a theoretically-derived robustness-accuracy tradeoff of the class conditional Gaussian mixture. Given a pretrained model, the representations of data synthesized from the Gaussian mixture are used to compare with our reference to infer the quality. By comparing the ratio of area-under-curve between the raw data and their representations, SynBench offers a quantifiable score for robustness-accuracy performance benchmarking. Our framework applies to a wide range of pretrained models taking continuous data inputs and is independent of the downstream tasks and datasets. Evaluated with several pretrained vision transformer models, the experimental results show that our SynBench score well matches the actual linear probing performance of the pre-trained model when fine-tuned on downstream tasks. Moreover, our framework can be used to inform the design of robust linear probing on pretrained representations to mitigate the robustness-accuracy tradeoff in downstream tasks. △ Less

Submitted 7 October, 2022; v1 submitted 6 October, 2022; originally announced October 2022.

Visual Pre-training for Navigation: What Can We Learn from Noise?

Authors: Yanwei Wang, Ching-Yun Ko, Pulkit Agrawal

Abstract: One powerful paradigm in visual navigation is to predict actions from observations directly. Training such an end-to-end system allows representations useful for downstream tasks to emerge automatically. However, the lack of inductive bias makes this system data inefficient. We hypothesize a sufficient representation of the current view and the goal view for a navigation policy can be learned by p… ▽ More One powerful paradigm in visual navigation is to predict actions from observations directly. Training such an end-to-end system allows representations useful for downstream tasks to emerge automatically. However, the lack of inductive bias makes this system data inefficient. We hypothesize a sufficient representation of the current view and the goal view for a navigation policy can be learned by predicting the location and size of a crop of the current view that corresponds to the goal. We further show that training such random crop prediction in a self-supervised fashion purely on synthetic noise images transfers well to natural home images. The learned representation can then be bootstrapped to learn a navigation policy efficiently with little interaction data. The code is available at https://yanweiw.github.io/noise2ptz △ Less

Submitted 26 July, 2023; v1 submitted 30 June, 2022; originally announced July 2022.

Comments: IROS 2023

Cyclops: Open Platform for Scale Truck Platooning

Authors: Hyeongyu Lee, Jaegeun Park, Changjin Koo, Jong-Chan Kim, Yongsoon Eun

Abstract: Cyclops, introduced in this paper, is an open research platform for everyone that wants to validate novel ideas and approaches in the area of self-driving heavy-duty vehicle platooning. The platform consists of multiple 1/14 scale semi-trailer trucks, a scale proving ground, and associated computing, communication and control modules that enable self-driving on the proving ground. A perception sys… ▽ More Cyclops, introduced in this paper, is an open research platform for everyone that wants to validate novel ideas and approaches in the area of self-driving heavy-duty vehicle platooning. The platform consists of multiple 1/14 scale semi-trailer trucks, a scale proving ground, and associated computing, communication and control modules that enable self-driving on the proving ground. A perception system for each vehicle is composed of a lidar-based object tracking system and a lane detection/control system. The former is to maintain the gap to the leading vehicle and the latter is to maintain the vehicle within the lane by steering control. The lane detection system is optimized for truck platooning where the field of view of the front-facing camera is severely limited due to a small gap to the leading vehicle. This platform is particularly amenable to validate mitigation strategies for safety-critical situations. Indeed, a simplex structure is adopted in the embedded module for testing various fail safe operations. We illustrate a scenario where camera sensor fails in the perception system but the vehicle operates at a reduced capacity to a graceful stop. Details of the Cyclops including 3D CAD designs and algorithm source codes are released for those who want to build similar testbeds. △ Less

Submitted 2 March, 2022; v1 submitted 10 February, 2022; originally announced February 2022.

Epistemic AI platform accelerates innovation by connecting biomedical knowledge

Authors: Da Chen Emily Koo, Heather Bowling, Kenneth Ashworth, David J. Heeger, Stefano Pacifico

Abstract: Epistemic AI accelerates biomedical discovery by finding hidden connections in the network of biomedical knowledge. The Epistemic AI web-based software platform embodies the concept of knowledge mapping, an interactive process that relies on a knowledge graph in combination with natural language processing (NLP), information retrieval, relevance feedback, and network analysis. Knowledge mapping re… ▽ More Epistemic AI accelerates biomedical discovery by finding hidden connections in the network of biomedical knowledge. The Epistemic AI web-based software platform embodies the concept of knowledge mapping, an interactive process that relies on a knowledge graph in combination with natural language processing (NLP), information retrieval, relevance feedback, and network analysis. Knowledge mapping reduces information overload, prevents costly mistakes, and minimizes missed opportunities in the research process. The platform combines state-of-the-art methods for information extraction with machine learning, artificial intelligence and network analysis. Starting from a single biological entity, such as a gene or disease, users may: a) construct a map of connections to that entity, b) map an entire domain of interest, and c) gain insight into large biological networks of knowledge. Knowledge maps provide clarity and organization, simplifying the day-to-day research processes. △ Less

Submitted 31 March, 2022; v1 submitted 27 January, 2022; originally announced January 2022.

Comments: 12 pages, 2 main figures

Revisiting Contrastive Learning through the Lens of Neighborhood Component Analysis: an Integrated Framework

Authors: Ching-Yun Ko, Jeet Mohapatra, Sijia Liu, Pin-Yu Chen, Luca Daniel, Lily Weng

Abstract: As a seminal tool in self-supervised representation learning, contrastive learning has gained unprecedented attention in recent years. In essence, contrastive learning aims to leverage pairs of positive and negative samples for representation learning, which relates to exploiting neighborhood information in a feature space. By investigating the connection between contrastive learning and neighborh… ▽ More As a seminal tool in self-supervised representation learning, contrastive learning has gained unprecedented attention in recent years. In essence, contrastive learning aims to leverage pairs of positive and negative samples for representation learning, which relates to exploiting neighborhood information in a feature space. By investigating the connection between contrastive learning and neighborhood component analysis (NCA), we provide a novel stochastic nearest neighbor viewpoint of contrastive learning and subsequently propose a series of contrastive losses that outperform the existing ones. Under our proposed framework, we show a new methodology to design integrated contrastive losses that could simultaneously achieve good accuracy and robustness on downstream tasks. With the integrated framework, we achieve up to 6\% improvement on the standard accuracy and 17\% improvement on the robust accuracy. △ Less

Submitted 28 January, 2022; v1 submitted 8 December, 2021; originally announced December 2021.

Two-Stage Mesh Deep Learning for Automated Tooth Segmentation and Landmark Localization on 3D Intraoral Scans

Authors: Tai-Hsien Wu, Chunfeng Lian, Sanghee Lee, Matthew Pastewait, Christian Piers, Jie Liu, Fang Wang, Li Wang, Chiung-Ying Chiu, Wenchi Wang, Christina Jackson, Wei-Lun Chao, Dinggang Shen, Ching-Chang Ko

Abstract: Accurately segmenting teeth and identifying the corresponding anatomical landmarks on dental mesh models are essential in computer-aided orthodontic treatment. Manually performing these two tasks is time-consuming, tedious, and, more importantly, highly dependent on orthodontists' experiences due to the abnormality and large-scale variance of patients' teeth. Some machine learning-based methods ha… ▽ More Accurately segmenting teeth and identifying the corresponding anatomical landmarks on dental mesh models are essential in computer-aided orthodontic treatment. Manually performing these two tasks is time-consuming, tedious, and, more importantly, highly dependent on orthodontists' experiences due to the abnormality and large-scale variance of patients' teeth. Some machine learning-based methods have been designed and applied in the orthodontic field to automatically segment dental meshes (e.g., intraoral scans). In contrast, the number of studies on tooth landmark localization is still limited. This paper proposes a two-stage framework based on mesh deep learning (called TS-MDL) for joint tooth labeling and landmark identification on raw intraoral scans. Our TS-MDL first adopts an end-to-end \emph{i}MeshSegNet method (i.e., a variant of the existing MeshSegNet with both improved accuracy and efficiency) to label each tooth on the downsampled scan. Guided by the segmentation outputs, our TS-MDL further selects each tooth's region of interest (ROI) on the original mesh to construct a light-weight variant of the pioneering PointNet (i.e., PointNet-Reg) for regressing the corresponding landmark heatmaps. Our TS-MDL was evaluated on a real-clinical dataset, showing promising segmentation and localization performance. Specifically, \emph{i}MeshSegNet in the first stage of TS-MDL reached an averaged Dice similarity coefficient (DSC) at \textcolor[rgb]{0,0,0}{$0.964\pm0.054$}, significantly outperforming the original MeshSegNet. In the second stage, PointNet-Reg achieved a mean absolute error (MAE) of $0.597\pm0.761 \, mm$ in distances between the prediction and ground truth for $66$ landmarks, which is superior compared with other networks for landmark detection. All these results suggest the potential usage of our TS-MDL in orthodontics. △ Less

Submitted 2 June, 2022; v1 submitted 24 September, 2021; originally announced September 2021.

Comments: 9 pages, 8 figures, accepted by IEEE TMI

A Meta-model for Process Failure Mode and Effects Analysis (PFMEA)

Authors: Kai Hoefig, Cornel Klein, Stefan Rothbauer, Marc Zeller, Marian Vorderer, Chee Hung Koo

Abstract: Short product lifecycles and a high variety of products force industrial manufacturing processes to change frequently. Due to the manual approach of many quality analysis techniques, they can significantly slow down adaption processes of production systems or make production unprofitable. Therefore, automating them can be a key technology for keeping pace with market demand of the future. The meth… ▽ More Short product lifecycles and a high variety of products force industrial manufacturing processes to change frequently. Due to the manual approach of many quality analysis techniques, they can significantly slow down adaption processes of production systems or make production unprofitable. Therefore, automating them can be a key technology for keeping pace with market demand of the future. The methodology presented here aims at a meta-model supporting automation for PFMEA. The method differentiates product requirements, production steps and quality measures in such a way, that complex quality requirements can be addressed in any instance of a factory using a common meta-modeling language. △ Less

Submitted 31 May, 2021; originally announced June 2021.

Comments: arXiv admin note: text overlap with arXiv:2105.14817

Journal ref: 2019 24th IEEE International Conference on Emerging Technologies and Factory Automation (ETFA)

SQUADfps: Integrated Model-Based Machine Safety and Product Quality for Flexible Production Systems

Authors: Chee Hung Koo, Stefan Rothbauer, Marian Vorderer, Kai Hoefig, Marc Zeller

Abstract: Growing individualization of products up to lot-size-1 and high volatility of product mixes lead to new challenges in the manufacturing domain, including the need for frequent reconfiguration of the system and reacting to changing orders. Thus, apart from functional aspects, safety aspects of the production system as well as product quality assurance aspects must be addressed for flexible and reco… ▽ More Growing individualization of products up to lot-size-1 and high volatility of product mixes lead to new challenges in the manufacturing domain, including the need for frequent reconfiguration of the system and reacting to changing orders. Thus, apart from functional aspects, safety aspects of the production system as well as product quality assurance aspects must be addressed for flexible and reconfigurable manufacturing systems at runtime. To cope with the mentioned challenges, we present an integrated model-based approach SQUADfps (machine Safety and product QUAlity for flexible proDuction systems) to support the automatic conduct of the risk assessment of flexible production scenarios in terms of safety as well as the process-FMEA to ensure that the requirements w.r.t. the quality of the production process and the resulting product are met. Our approach is based on a meta-model which captures all information needed to conduct both risk assessment and process-FMEA dynamically during the runtime, and thus enables flexible manufacturing scenarios with frequent changes of the production system and orders up to a lot-size of one while guaranteeing safety and product quality requirements. The automatically generated results will assist human in making further decisions. To demonstrate the feasibility of our approach, we apply it to a case study. △ Less

Submitted 4 June, 2021; v1 submitted 31 May, 2021; originally announced May 2021.

Journal ref: Papadopoulos Y., Aslansefat K., Katsaros P., Bozzano M. (eds) Model-Based Safety and Assessment. IMBSA 2019. Lecture Notes in Computer Science, vol 11842. Springer, Cham

MoSES_2PDF: A GIS-Compatible GPU-accelerated High-Performance Simulation Tool for Grain-Fluid Shallow Flows

Authors: Chi-Jyun Ko, Po-Chih Chen, Hock-Kiet Wong, Yih-Chin Tai

Abstract: We introduce a GPU-accelerated simulation tool, named Modeling on Shallow Flows with Efficient Simulation for Two-Phase Debris Flows (MoSES_2PDF), of which the input and output data can be linked to the GIS system for engineering application. MoSES_2PDF is developed based on the CUDA structure so that it can well run with different NVIDIA GPU cards, once the CUDA vers. 9.2 (or higher) is installed… ▽ More We introduce a GPU-accelerated simulation tool, named Modeling on Shallow Flows with Efficient Simulation for Two-Phase Debris Flows (MoSES_2PDF), of which the input and output data can be linked to the GIS system for engineering application. MoSES_2PDF is developed based on the CUDA structure so that it can well run with different NVIDIA GPU cards, once the CUDA vers. 9.2 (or higher) is installed. The performance of the MoSES_2PDF is evaluated, and it is found that the present GPU-CUDA implementation can enhance efficiency by up to 230 folds, depending on the PC/workstations, models of GPU card, and the mesh numbers in the computation domain. Two numerical examples are illustrated with two distinct initial inflow conditions, which are included in two modes of MoSES_2PDF, respectively. In the numerical example of a large-scale event, the 2009 Hsiaolin event, the results computed by two distinct NVIDIA GPU cards (RTX-2080-Ti and Tesla-V100) are found to be identical but only tiny deviation is figured out in comparison with the results computed by the conventional single-core CPU-code. It is speculated to be caused by the different structures in the source codes and some float/double operations. In addition to the illustration in the GIS system, the computed results by MoSES\_2PDF can also be shown with animated 3D graphics in the ANSI-Platform, where the user can interact with 3D scenes. The feasibility, features, and facilities of MoSES\_2PDF are demonstrated with respect to the two numerical examples concerning two real events. △ Less

Submitted 14 April, 2021; originally announced April 2021.

Comments: 16 pages, 7 figures and 1 table

A Deep Learning Approach for Characterizing Major Galaxy Mergers

Authors: Skanda Koppula, Victor Bapst, Marc Huertas-Company, Sam Blackwell, Agnieszka Grabska-Barwinska, Sander Dieleman, Andrea Huber, Natasha Antropova, Mikolaj Binkowski, Hannah Openshaw, Adria Recasens, Fernando Caro, Avishai Deke, Yohan Dubois, Jesus Vega Ferrero, David C. Koo, Joel R. Primack, Trevor Back

Abstract: Fine-grained estimation of galaxy merger stages from observations is a key problem useful for validation of our current theoretical understanding of galaxy formation. To this end, we demonstrate a CNN-based regression model that is able to predict, for the first time, using a single image, the merger stage relative to the first perigee passage with a median error of 38.3 million years (Myrs) over… ▽ More Fine-grained estimation of galaxy merger stages from observations is a key problem useful for validation of our current theoretical understanding of galaxy formation. To this end, we demonstrate a CNN-based regression model that is able to predict, for the first time, using a single image, the merger stage relative to the first perigee passage with a median error of 38.3 million years (Myrs) over a period of 400 Myrs. This model uses no specific dynamical modeling and learns only from simulated merger events. We show that our model provides reasonable estimates on real observations, approximately matching prior estimates provided by detailed dynamical modeling. We provide a preliminary interpretability analysis of our models, and demonstrate first steps toward calibrated uncertainty estimation. △ Less

Submitted 9 February, 2021; originally announced February 2021.

Comments: Third Workshop on Machine Learning and the Physical Sciences (NeurIPS 2020), Vancouver, Canada

Higher-Order Certification for Randomized Smoothing

Authors: Jeet Mohapatra, Ching-Yun Ko, Tsui-Wei Weng, Pin-Yu Chen, Sijia Liu, Luca Daniel

Abstract: Randomized smoothing is a recently proposed defense against adversarial attacks that has achieved SOTA provable robustness against $\ell_2$ perturbations. A number of publications have extended the guarantees to other metrics, such as $\ell_1$ or $\ell_\infty$, by using different smoothing measures. Although the current framework has been shown to yield near-optimal $\ell_p$ radii, the total safet… ▽ More Randomized smoothing is a recently proposed defense against adversarial attacks that has achieved SOTA provable robustness against $\ell_2$ perturbations. A number of publications have extended the guarantees to other metrics, such as $\ell_1$ or $\ell_\infty$, by using different smoothing measures. Although the current framework has been shown to yield near-optimal $\ell_p$ radii, the total safety region certified by the current framework can be arbitrarily small compared to the optimal. In this work, we propose a framework to improve the certified safety region for these smoothed classifiers without changing the underlying smoothing scheme. The theoretical contributions are as follows: 1) We generalize the certification for randomized smoothing by reformulating certified radius calculation as a nested optimization problem over a class of functions. 2) We provide a method to calculate the certified safety region using $0^{th}$-order and $1^{st}$-order information for Gaussian-smoothed classifiers. We also provide a framework that generalizes the calculation for certification using higher-order information. 3) We design efficient, high-confidence estimators for the relevant statistics of the first-order information. Combining the theoretical contribution 2) and 3) allows us to certify safety region that are significantly larger than the ones provided by the current methods. On CIFAR10 and Imagenet datasets, the new regions certified by our approach achieve significant improvements on general $\ell_1$ certified radii and on the $\ell_2$ certified radii for color-space attacks ($\ell_2$ restricted to 1 channel) while also achieving smaller improvements on the general $\ell_2$ certified radii. Our framework can also provide a way to circumvent the current impossibility results on achieving higher magnitude of certified radii without requiring the use of data-dependent smoothing techniques. △ Less

Submitted 13 October, 2020; originally announced October 2020.

Comments: Accepted to NeurIPS2020(spotlight)

Hidden Cost of Randomized Smoothing

Authors: Jeet Mohapatra, Ching-Yun Ko, Tsui-Wei, Weng, Sijia Liu, Pin-Yu Chen, Luca Daniel

Abstract: The fragility of modern machine learning models has drawn a considerable amount of attention from both academia and the public. While immense interests were in either crafting adversarial attacks as a way to measure the robustness of neural networks or devising worst-case analytical robustness verification with guarantees, few methods could enjoy both scalability and robustness guarantees at the s… ▽ More The fragility of modern machine learning models has drawn a considerable amount of attention from both academia and the public. While immense interests were in either crafting adversarial attacks as a way to measure the robustness of neural networks or devising worst-case analytical robustness verification with guarantees, few methods could enjoy both scalability and robustness guarantees at the same time. As an alternative to these attempts, randomized smoothing adopts a different prediction rule that enables statistical robustness arguments which easily scale to large networks. However, in this paper, we point out the side effects of current randomized smoothing workflows. Specifically, we articulate and prove two major points: 1) the decision boundaries of smoothed classifiers will shrink, resulting in disparity in class-wise accuracy; 2) applying noise augmentation in the training process does not necessarily resolve the shrinking issue due to the inconsistent learning objectives. △ Less

Submitted 12 March, 2021; v1 submitted 2 March, 2020; originally announced March 2020.

Comments: Jeet Mohapatra and Ching-Yun Ko contributed equally. To appear in AISTATS 2021

HOTCAKE: Higher Order Tucker Articulated Kernels for Deeper CNN Compression

Authors: Rui Lin, Ching-Yun Ko, Zhuolun He, Cong Chen, Yuan Cheng, Hao Yu, Graziano Chesi, Ngai Wong

Abstract: The emerging edge computing has promoted immense interests in compacting a neural network without sacrificing much accuracy. In this regard, low-rank tensor decomposition constitutes a powerful tool to compress convolutional neural networks (CNNs) by decomposing the 4-way kernel tensor into multi-stage smaller ones. Building on top of Tucker-2 decomposition, we propose a generalized Higher Order T… ▽ More The emerging edge computing has promoted immense interests in compacting a neural network without sacrificing much accuracy. In this regard, low-rank tensor decomposition constitutes a powerful tool to compress convolutional neural networks (CNNs) by decomposing the 4-way kernel tensor into multi-stage smaller ones. Building on top of Tucker-2 decomposition, we propose a generalized Higher Order Tucker Articulated Kernels (HOTCAKE) scheme comprising four steps: input channel decomposition, guided Tucker rank selection, higher order Tucker decomposition and fine-tuning. By subjecting each CONV layer to HOTCAKE, a highly compressed CNN model with graceful accuracy trade-off is obtained. Experiments show HOTCAKE can compress even pre-compressed models and produce state-of-the-art lightweight networks. △ Less

Submitted 28 February, 2020; originally announced February 2020.

Comments: 6 pages, 5 figures

Interpretation and Simplification of Deep Forest

Authors: Sangwon Kim, Mira Jeong, Byoung Chul Ko

Abstract: This paper proposes a new method for interpreting and simplifying a black box model of a deep random forest (RF) using a proposed rule elimination. In deep RF, a large number of decision trees are connected to multiple layers, thereby making an analysis difficult. It has a high performance similar to that of a deep neural network (DNN), but achieves a better generalizability. Therefore, in this st… ▽ More This paper proposes a new method for interpreting and simplifying a black box model of a deep random forest (RF) using a proposed rule elimination. In deep RF, a large number of decision trees are connected to multiple layers, thereby making an analysis difficult. It has a high performance similar to that of a deep neural network (DNN), but achieves a better generalizability. Therefore, in this study, we consider quantifying the feature contributions and frequency of the fully trained deep RF in the form of a decision rule set. The feature contributions provide a basis for determining how features affect the decision process in a rule set. Model simplification is achieved by eliminating unnecessary rules by measuring the feature contributions. Consequently, the simplified model has fewer parameters and rules than before. Experiment results have shown that a feature contribution analysis allows a black box model to be decomposed for quantitatively interpreting a rule set. The proposed method was successfully applied to various deep RF models and benchmark datasets while maintaining a robust performance despite the elimination of a large number of rules. △ Less

Submitted 11 December, 2020; v1 submitted 14 January, 2020; originally announced January 2020.

Comments: Major issues on the experiments

Fastened CROWN: Tightened Neural Network Robustness Certificates

Authors: Zhaoyang Lyu, Ching-Yun Ko, Zhifeng Kong, Ngai Wong, Dahua Lin, Luca Daniel

Abstract: The rapid growth of deep learning applications in real life is accompanied by severe safety concerns. To mitigate this uneasy phenomenon, much research has been done providing reliable evaluations of the fragility level in different deep neural networks. Apart from devising adversarial attacks, quantifiers that certify safeguarded regions have also been designed in the past five years. The summari… ▽ More The rapid growth of deep learning applications in real life is accompanied by severe safety concerns. To mitigate this uneasy phenomenon, much research has been done providing reliable evaluations of the fragility level in different deep neural networks. Apart from devising adversarial attacks, quantifiers that certify safeguarded regions have also been designed in the past five years. The summarizing work of Salman et al. unifies a family of existing verifiers under a convex relaxation framework. We draw inspiration from such work and further demonstrate the optimality of deterministic CROWN (Zhang et al. 2018) solutions in a given linear programming problem under mild constraints. Given this theoretical result, the computationally expensive linear programming based method is shown to be unnecessary. We then propose an optimization-based approach \textit{FROWN} (\textbf{F}astened C\textbf{ROWN}): a general algorithm to tighten robustness certificates for neural networks. Extensive experiments on various networks trained individually verify the effectiveness of FROWN in safeguarding larger robust regions. △ Less

Submitted 1 December, 2019; originally announced December 2019.

Comments: Zhaoyang Lyu and Ching-Yun Ko contributed equally, accepted to AAAI 2020