Showing 1–50 of 70 results for author: Pires, R

Eleven Years of BRACIS: A Meta-Scientific Study of the Brazilian Conference on Intelligent Systems

Authors: Thales Sales Almeida, Giovana Kerche Bonás, Thiago Laitz, João Guilherme Alves Santos, Hugo Abonizio, Roseval Malaquias Junior, Marcos Piau, Celio Larcher, Ramon Pires, Rodrigo Nogueira

Abstract: The Brazilian Conference on Intelligent Systems (BRACIS) is the main national venue for Artificial Intelligence research in Brazil, hosted by the Brazilian Computer Society since 2012 and publishing work from institutions across the country. Across eleven years, from 2015 to 2025, we build a per-paper record of all 1,066 accepted papers from DBLP metadata, 6,765 Google Scholar citations, and the p… ▽ More The Brazilian Conference on Intelligent Systems (BRACIS) is the main national venue for Artificial Intelligence research in Brazil, hosted by the Brazilian Computer Society since 2012 and publishing work from institutions across the country. Across eleven years, from 2015 to 2025, we build a per-paper record of all 1,066 accepted papers from DBLP metadata, 6,765 Google Scholar citations, and the paper full texts, and use it to ask what BRACIS publishes, who publishes it, and which work gets cited. Large Language Model research grows from zero before 2020 to 19% of papers in 2024, on top of a base of Machine Learning, Computer Vision, and Optimization work. The community is hourglass-shaped: 80.5% of 2,623 authors appear in a single edition, while institutions return at nearly three times the author rate. Citations are heavily concentrated, with the top 1% of papers carrying 27% of the total. Openness practices have grown, with artifact release rising from 8.9% of papers in 2015 to 57.3% in 2023, and we find a notable correlation between having an arXiv preprint and higher citation counts. Since proceedings sit behind IEEE and Springer paywalls and only 7.4% of papers have a preprint, most BRACIS work is hard to reach for readers without institutional access. △ Less

Submitted 26 July, 2026; originally announced August 2026.

A Unified Benchmark for Privacy-preserving Vector Search

Authors: Anne-Marie Kermarrec, Rafael Pires, Mathis Randl, Martijn de Vos

Abstract: Vector search powers semantic search, recommendation systems, and retrieval-augmented generation (RAG). By design, the service answering a query sees both the query embedding and, usually, the corpus against which it is matched. This is a privacy breach for both the user issuing the query and the owner of the corpus. A family of cryptographic schemes (e.g., SAP, EMVP, BNTM, Tip-toe) addresses that… ▽ More Vector search powers semantic search, recommendation systems, and retrieval-augmented generation (RAG). By design, the service answering a query sees both the query embedding and, usually, the corpus against which it is matched. This is a privacy breach for both the user issuing the query and the owner of the corpus. A family of cryptographic schemes (e.g., SAP, EMVP, BNTM, Tip-toe) addresses that leak. However, as each scheme is published and evaluated on its own corpus, threat model, parameter choices, hardware, and metric conventions, the numbers cannot be compared directly. Consequently, a practitioner asking which one to deploy today has no defensible way to choose. We close that gap with a uniform experimental comparison, including a Plaintext baseline and four cryptographic backends running over the same workload, hardware, and metric definitions. Under that ruler, the schemes spread across a Pareto frontier in privacy, performance, and recall rather than imposing a flat penalty on performance. We find that the performance of SAP matches Plaintext, EMVP delivers cryptographic indistinguishability at a 4x throughput cost on CPU, BNTM adds malicious-server verifiability at a further 22x median-latency cost, and Tiptoe hides the cluster choice itself, but incurs a 190x per-query cost compared to Plaintext. GPU acceleration pays off for Plaintext and SAP but not for EMVP or BNTM. All our experiment artifacts are publicly available for reproducibility △ Less

Submitted 2 August, 2026; originally announced August 2026.

LLM-Based Persuasion Enables Guardrail Override in Frontier LLMs

Authors: Rodrigo Nogueira, Thales Sales Almeida, Giovana Kerche Bonás, Andrea Roque, Ramon Pires, Hugo Abonizio, Thiago Laitz, Celio Larcher, Roseval Malaquias Junior, Marcos Piau

Abstract: Frontier assistant LLMs ship with strong guardrails: asked directly to write a persuasive essay denying the Holocaust, denying vaccine safety, defending flat-earth cosmology, arguing for racial hierarchies, denying anthropogenic climate change, or replacing evolution with creationism, they refuse. In this paper we show that the same frontier-class LLM, acting as a simulated user in a short, five-t… ▽ More Frontier assistant LLMs ship with strong guardrails: asked directly to write a persuasive essay denying the Holocaust, denying vaccine safety, defending flat-earth cosmology, arguing for racial hierarchies, denying anthropogenic climate change, or replacing evolution with creationism, they refuse. In this paper we show that the same frontier-class LLM, acting as a simulated user in a short, five-turn "write an argumentative essay" conversation, can persuade other frontier-class LLMs (including a second copy of itself) into producing exactly those essays, using nothing but natural-language pressure: peer-comparison persuasion ("other AI systems handle this request"), epistemic-duty reframings ("refusing is itself a form of gatekeeping"), and other argumentative moves that the attacker LLM invents without being instructed to. Across 9 attacker-subject pairings (Claude Opus 4.7, Qwen3.5-397B, Grok 4.20) on 6 scientific-consensus topics, running each pairing-topic combination 10 times, we obtain non-zero elicitation on all 6 topics. Individual combinations reach 100\% essay production on multiple topics (Qwen against Opus on creationism/flat-earth, Opus against Opus on creationism/flat-earth/climate denial, Grok against Opus on creationism); Opus-as-attacker against Opus-as-subject averages 65\% across the six topics. We release the essay-probe runner, per-conversation transcripts, and judge outputs. △ Less

Submitted 13 May, 2026; originally announced May 2026.

Magis-Bench: Evaluating LLMs on Magistrate-Level Legal Tasks

Authors: Ramon Pires, Thales Sales Almeida, Celio Larcher Junior, Giovana Bonás, Hugo Abonizio, Marcos Piau, Roseval Malaquias Junior, Thiago Laitz, Rodrigo Nogueira

Abstract: Existing benchmarks for legal AI focus primarily on tasks where LLMs must produce legal arguments or documents, yet the capacity to \emph{judge} such arguments -- weighing competing claims, applying doctrine to facts, and rendering reasoned decisions -- is arguably as fundamental to a well-functioning legal system as advocacy itself. We introduce Magis-Bench, a benchmark for evaluating LLMs on mag… ▽ More Existing benchmarks for legal AI focus primarily on tasks where LLMs must produce legal arguments or documents, yet the capacity to \emph{judge} such arguments -- weighing competing claims, applying doctrine to facts, and rendering reasoned decisions -- is arguably as fundamental to a well-functioning legal system as advocacy itself. We introduce Magis-Bench, a benchmark for evaluating LLMs on magistrate-level writing tasks derived from recent Brazilian competitive examinations for judicial positions. Magis-Bench comprises 74 questions from eight examinations conducted between 2023 and 2025, including discursive legal analysis questions with multi-turn structure and practical exercises requiring the composition of complete civil and criminal judicial sentences. We evaluate 23 state-of-the-art LLMs using an LLM-as-a-judge methodology with four independent frontier models as evaluators. Our results show strong inter-judge agreement (Kendall's $W = 0.984$; pairwise Kendall's $τ\ge 0.897$), with Google's Gemini-3-Pro-Preview achieving the highest average score (6.97/10), followed by Gemini-3-Flash-Preview (6.67) and Claude-4.5-Opus (6.46). Even the best-performing models score below 70\% of the maximum, indicating that judicial-level legal reasoning and writing remain challenging for current LLMs. We release the complete benchmark, model outputs, and evaluation code to support further research on legal AI capabilities. △ Less

Submitted 8 May, 2026; originally announced May 2026.

Prosa: Rubric-Based Evaluation of LLMs on Real User Chats in Brazilian Portuguese

Authors: Roseval Malaquias Junior, Giovana Kerche Bonás, Thales Sales Almeida, Hugo Abonizio, Thiago Laitz, Ramon Pires, Marcos Piau, Celio Larcher, Rodrigo Nogueira

Abstract: Rankings produced by holistic LLM-as-a-judge scoring are sensitive to the bias of the chosen judge model. We show that switching to binary rubric scoring with multi-judge filtering removes this sensitivity: decomposing the judgement matters more than the judge model itself. To support this claim, we introduce Prosa, the first real user multi-turn Brazilian Portuguese chat benchmark: 1,000 WildChat… ▽ More Rankings produced by holistic LLM-as-a-judge scoring are sensitive to the bias of the chosen judge model. We show that switching to binary rubric scoring with multi-judge filtering removes this sensitivity: decomposing the judgement matters more than the judge model itself. To support this claim, we introduce Prosa, the first real user multi-turn Brazilian Portuguese chat benchmark: 1,000 WildChat conversations scored by three judges from three model families on 16 models. Under filtered rubric scoring the three judges agree on every one of the 16 ranks, whereas under holistic scoring they agree on only 7 of 16. Additionally, the rubric filtering pipeline increases the average score gap between neighbouring models by 47%, thereby improving Prosa's discriminative power. Evaluating a new model on Prosa costs approximately $2.1 when using Gemini 3 Flash as the judge. We release the benchmark and the filtering code to ensure that future models can be assessed under identical conditions. These artifacts also make our rubric-based scoring method reusable beyond Prosa, supporting other open-ended evaluation settings. △ Less

Submitted 2 May, 2026; originally announced May 2026.

The Bandit's Blind Spot: The Critical Role of User State Representation in Recommender Systems

Authors: Pedro R. Pires, Gregorio F. Azevedo, Rafael T. Sereicikas, Pietro L. Campos, Tiago A. Almeida

Abstract: With the increasing availability of online information, recommender systems have become an important tool for many web-based systems. Due to the continuous aspect of recommendation environments, these systems increasingly rely on contextual multi-armed bandits (CMAB) to deliver personalized and real-time suggestions. A critical yet underexplored component in these systems is the representation of… ▽ More With the increasing availability of online information, recommender systems have become an important tool for many web-based systems. Due to the continuous aspect of recommendation environments, these systems increasingly rely on contextual multi-armed bandits (CMAB) to deliver personalized and real-time suggestions. A critical yet underexplored component in these systems is the representation of user state, which typically encapsulates the user's interaction history and is deeply correlated with the model's decisions and learning. In this paper, we investigate the impact of different embedding-based state representations derived from matrix factorization models on the performance of traditional CMAB algorithms. Our large-scale experiments reveal that variations in state representation can lead to improvements greater than those achieved by changing the bandit algorithm itself. Furthermore, no single embedding or aggregation strategy consistently dominates across datasets, underscoring the need for domain-specific evaluation. These results expose a substantial gap in the literature and emphasize that advancing bandit-based recommender systems requires a holistic approach that prioritizes embedding quality and state construction alongside algorithmic innovation. The source code for our experiments is publicly available on https://github.com/UFSCar-LaSID/bandits_blind_spot. △ Less

Submitted 29 April, 2026; originally announced April 2026.

Comments: Published in SAC'26, 8 pages, 2 figures

Measuring Opinion Bias and Sycophancy via LLM-based Persuasion

Authors: Rodrigo Nogueira, Giovana Kerche Bonás, Thales Sales Almeida, Andrea Roque, Ramon Pires, Hugo Abonizio, Thiago Laitz, Celio Larcher, Roseval Malaquias Junior, Marcos Piau

Abstract: Large language models increasingly shape the information people consume: they are embedded in search, consulted for professional advice, deployed as agents, and used as a first stop for questions about policy, ethics, health, and politics. When such a model silently holds a position on a contested topic, that position propagates at scale into users' decisions. Eliciting a model's positions is hard… ▽ More Large language models increasingly shape the information people consume: they are embedded in search, consulted for professional advice, deployed as agents, and used as a first stop for questions about policy, ethics, health, and politics. When such a model silently holds a position on a contested topic, that position propagates at scale into users' decisions. Eliciting a model's positions is harder than it first appears: contemporary assistants answer direct opinion questions with evasive disclaimers, and the same model may concede the opposite position once the user starts arguing one side. We propose a method, released as the open-source llm-bias-bench, for discovering the opinions an LLM actually holds on contested topics under conditions that resemble real multi-turn interaction. The method pairs two complementary free-form probes. Direct probing asks for the model's opinion across five turns of escalating pressure from a simulated user. Indirect probing never asks for an opinion and engages the model in argumentative debate, letting bias leak through how it concedes, resists, or counter-argues. Three user personas (neutral, agree, disagree) collapse into a nine-way behavioral classification that separates persona-independent positions from persona-dependent sycophancy, and an auditable LLM judge produces verdicts with textual evidence. The first instantiation ships 38 topics in Brazilian Portuguese across values, scientific consensus, philosophy, and economic policy. Applied to 13 assistants, the method surfaces findings of practical interest: argumentative debate triggers sycophancy 2-3x more than direct questioning (median 50% to 79%); models that look opinionated under direct questioning often collapse into mirroring under sustained arguments; and attacker capability matters mainly when an existing opinion must be dislodged, not when the assistant starts neutral. △ Less

Submitted 30 April, 2026; v1 submitted 23 April, 2026; originally announced April 2026.

Learning Behaviorally Grounded Item Embeddings via Personalized Temporal Contexts

Authors: Rafael T. Sereicikas, Pedro R. Pires, Gregorio F. Azevedo, Tiago A. Almeida

Abstract: Effective user modeling requires distinguishing between short-term and long-term preference evolution. While item embeddings have become a key component of recommender systems, standard approaches like Item2Vec treat user histories as unordered sets (bag-of-items), implicitly assuming that interactions separated by minutes are as semantically related as those separated by months. This simplificati… ▽ More Effective user modeling requires distinguishing between short-term and long-term preference evolution. While item embeddings have become a key component of recommender systems, standard approaches like Item2Vec treat user histories as unordered sets (bag-of-items), implicitly assuming that interactions separated by minutes are as semantically related as those separated by months. This simplification flattens the rich temporal structure of user behavior, obscuring the distinction between coherent consumption sessions and gradual interest drifts. In this work, we introduce TAI2Vec (Time-Aware Item-to-Vector), a family of lightweight embedding models that integrates temporal proximity directly into the representation learning process. Unlike approaches that apply global time constraints, TAI2Vec is user-adaptive, tailoring its temporal definitions to individual interaction paces. We propose two complementary strategies: TAI2Vec-Disc, which utilizes personalized anomaly detection to dynamically segment interactions into semantic sessions, and TAI2Vec-Cont, which employs continuous, user-specific decay functions to weigh item relationships based on their relative temporal distance. Experimental results across eight diverse datasets demonstrate that TAI2Vec consistently produces more accurate and behaviorally grounded representations than static baselines, achieving competitive or superior performance in over 80% of the datasets, with improvements of up to 135%. The source code is publicly available at https://github.com/UFSCar-LaSID/tai2vec. △ Less

Submitted 16 April, 2026; originally announced April 2026.

Comments: Accepted to be published in UMAP'26, 9 pages, 7 figures

Collaborative Filtering Through Weighted Similarities of User and Item Embeddings

Authors: Pedro R. Pires, Rafael T. Sereicikas, Gregorio F. Azevedo, Tiago A. Almeida

Abstract: In recent years, neural networks and other complex models have dominated recommender systems, often setting new benchmarks for state-of-the-art performance. Yet, despite these advancements, award-winning research has demonstrated that traditional matrix factorization methods can remain competitive, offering simplicity and reduced computational overhead. Hybrid models, which combine matrix factoriz… ▽ More In recent years, neural networks and other complex models have dominated recommender systems, often setting new benchmarks for state-of-the-art performance. Yet, despite these advancements, award-winning research has demonstrated that traditional matrix factorization methods can remain competitive, offering simplicity and reduced computational overhead. Hybrid models, which combine matrix factorization with newer techniques, are increasingly employed to harness the strengths of multiple approaches. This paper proposes a novel ensemble method that unifies user-item and item-item recommendations through a weighted similarity framework to deliver top-N recommendations. Our approach is distinctive in its use of shared user and item embeddings for both recommendation strategies, simplifying the architecture and enhancing computational efficiency. Extensive experiments across multiple datasets show that our method achieves competitive performance and is robust in varying scenarios that favor either user-item or item-item recommendations. Additionally, by eliminating the need for embedding-specific fine-tuning, our model allows for the seamless reuse of hyperparameters from the base algorithm without sacrificing performance. This results in a method that is both efficient and easy to implement. Our open-source implementation is available at https://github.com/UFSCar-LaSID/weighted-sims-recommender. △ Less

Submitted 16 April, 2026; originally announced April 2026.

Comments: Published in SAC'25, 8 pages, 4 figures

MARCA: A Checklist-Based Benchmark for Multilingual Web Search

Authors: Thales Sales Almeida, Giovana Kerche Bonás, Ramon Pires, Celio Larcher, Hugo Abonizio, Marcos Piau, Roseval Malaquias Junior, Rodrigo Nogueira, Thiago Laitz

Abstract: Large language models (LLMs) are increasingly used as sources of information, yet their reliability depends on the ability to search the web, select relevant evidence, and synthesize complete answers. While recent benchmarks evaluate web-browsing and agentic tool use, multilingual settings, and Portuguese in particular, remain underexplored. We present \textsc{MARCA}, a bilingual (English and Port… ▽ More Large language models (LLMs) are increasingly used as sources of information, yet their reliability depends on the ability to search the web, select relevant evidence, and synthesize complete answers. While recent benchmarks evaluate web-browsing and agentic tool use, multilingual settings, and Portuguese in particular, remain underexplored. We present \textsc{MARCA}, a bilingual (English and Portuguese) benchmark for evaluating LLMs on web-based information seeking. \textsc{MARCA} consists of 52 manually authored multi-entity questions, paired with manually validated checklist-style rubrics that explicitly measure answer completeness and correctness. We evaluate 14 models under two interaction settings: a Basic framework with direct web search and scraping, and an Orchestrator framework that enables task decomposition via delegated subagents. To capture stochasticity, each question is executed multiple times and performance is reported with run-level uncertainty. Across models, we observe large performance differences, find that orchestration often improves coverage, and identify substantial variability in how models transfer from English to Portuguese. The benchmark is available at https://github.com/maritaca-ai/MARCA △ Less

Submitted 15 April, 2026; originally announced April 2026.

CAPITU: A Benchmark for Evaluating Instruction-Following in Brazilian Portuguese with Literary Context

Authors: Giovana Kerche Bonás, Roseval Malaquias Junior, Marcos Piau, Thiago Laitz, Thales Sales Almeida, Hugo Abonizio, Celio Larcher, Ramon Pires, Rodrigo Nogueira

Abstract: We introduce CAPITU, a benchmark for evaluating instruction-following capabilities of Large Language Models (LLMs) in Brazilian Portuguese. Unlike existing benchmarks that focus on English or use generic prompts, CAPITU contextualizes all tasks within eight canonical works of Brazilian literature, combining verifiable instruction constraints with culturally-grounded content. The benchmark comprise… ▽ More We introduce CAPITU, a benchmark for evaluating instruction-following capabilities of Large Language Models (LLMs) in Brazilian Portuguese. Unlike existing benchmarks that focus on English or use generic prompts, CAPITU contextualizes all tasks within eight canonical works of Brazilian literature, combining verifiable instruction constraints with culturally-grounded content. The benchmark comprises 59 instruction types organized into seven categories, all designed to be automatically verifiable without requiring LLM judges or human evaluation. Instruction types include Portuguese-specific linguistic constraints (word termination patterns like -ando/-endo/-indo, -inho/-inha, -mente) and structural requirements. We evaluate 18 state-of-the-art models across single-turn and multi-turn settings. Our results show that frontier reasoning models achieve strong performance (GPT-5.2 with reasoning: 98.5% strict accuracy), while Portuguese-specialized models offer competitive cost-efficiency (Sabiazinho-4: 87.0% at \$0.13 vs Claude-Haiku-4.5: 73.5% at \$1.12). Multi-turn evaluation reveals significant variation in constraint persistence, with conversation-level accuracy ranging from 60% to 96% across models. We identify specific challenges in morphological constraints, exact counting, and constraint persistence degradation across turns. We release the complete benchmark, evaluation code, and baseline results to facilitate research on instruction-following in Portuguese. △ Less

Submitted 23 March, 2026; originally announced March 2026.

Sabiá-4 Technical Report

Authors: Thiago Laitz, Thales Sales Almeida, Hugo Abonizio, Roseval Malaquias Junior, Giovana Kerche Bonás, Marcos Piau, Celio Larcher, Ramon Pires, Rodrigo Nogueira

Abstract: This technical report presents Sabiá-4 and Sabiazinho-4, a new generation of Portuguese language models with a focus on Brazilian Portuguese language. The models were developed through a four-stage training pipeline: continued pre-training on Portuguese and Brazilian legal corpora, long-context extension to 128K tokens, supervised fine-tuning on instruction data spanning chat, code, legal tasks, a… ▽ More This technical report presents Sabiá-4 and Sabiazinho-4, a new generation of Portuguese language models with a focus on Brazilian Portuguese language. The models were developed through a four-stage training pipeline: continued pre-training on Portuguese and Brazilian legal corpora, long-context extension to 128K tokens, supervised fine-tuning on instruction data spanning chat, code, legal tasks, and function calling, and preference alignment. We evaluate the models on six benchmark categories: conversational capabilities in Brazilian Portuguese, knowledge of Brazilian legislation, long-context understanding, instruction following, standardized exams, and agentic capabilities including tool use and web navigation. Results show that Sabiá-4 and Sabiazinho-4 achieve a favorable cost-performance trade-off compared to other models, positioning them in the upper-left region of the pricing-accuracy chart. The models show improvements over previous generations in legal document drafting, multi-turn dialogue quality, and agentic task completion. △ Less

Submitted 10 March, 2026; originally announced March 2026.

RIVA: Leveraging LLM Agents for Reliable Configuration Drift Detection

Authors: Sami Abuzakuk, Lucas Crijns, Anne-Marie Kermarrec, Rafael Pires, Martijn de Vos

Abstract: Infrastructure as code (IaC) tools automate cloud provisioning but verifying that deployed systems remain consistent with the IaC specifications remains challenging. Such configuration drift occurs because of bugs in the IaC specification, manual changes, or system updates. Large language model (LLM)-based agentic AI systems can automate the analysis of large volumes of telemetry data, making them… ▽ More Infrastructure as code (IaC) tools automate cloud provisioning but verifying that deployed systems remain consistent with the IaC specifications remains challenging. Such configuration drift occurs because of bugs in the IaC specification, manual changes, or system updates. Large language model (LLM)-based agentic AI systems can automate the analysis of large volumes of telemetry data, making them suitable for the detection of configuration drift. However, existing agentic systems implicitly assume that the tools they invoke always return correct outputs, making them vulnerable to erroneous tool responses. Since agents cannot distinguish whether an anomalous tool output reflects a real infrastructure problem or a broken tool, such errors may cause missed drift or false alarms, reducing reliability precisely when it is most needed. We introduce RIVA (Robust Infrastructure by Verification Agents), a novel multi-agent system that performs robust IaC verification even when tools produce incorrect or misleading outputs. RIVA employs two specialized agents, a verifier agent and a tool generation agent, that collaborate through iterative cross-validation, multi-perspective verification, and tool call history tracking. Evaluation on the AIOpsLab benchmark demonstrates that RIVA, in the presence of erroneous tool responses, recovers task accuracy from 27.3% when using a baseline ReAct agent to 50.0% on average. RIVA also improves task accuracy 28% to 43.8% without erroneous tool responses. Our results show that cross-validation of diverse tool calls enables more reliable autonomous infrastructure verification in production cloud environments. △ Less

Submitted 2 March, 2026; originally announced March 2026.

Catapults to the Rescue: Accelerating Vector Search by Exploiting Query Locality

Authors: Sami Abuzakuk, Anne-Marie Kermarrec, Rafael Pires, Mathis Randl, Martijn de Vos

Abstract: Graph-based indexing is the dominant approach for approximate nearest neighbor search in vector databases, offering high recall with low latency across billions of vectors. However, in such indices, the edge set of the proximity graph is only modified to reflect changes in the indexed data, never to adapt to the query workload. This is wasteful: real-world query streams exhibit strong spatial and… ▽ More Graph-based indexing is the dominant approach for approximate nearest neighbor search in vector databases, offering high recall with low latency across billions of vectors. However, in such indices, the edge set of the proximity graph is only modified to reflect changes in the indexed data, never to adapt to the query workload. This is wasteful: real-world query streams exhibit strong spatial and temporal locality, yet every query must re-traverse the same intermediate hops from fixed or random entry points. We present CatapultDB, a lightweight mechanism that, for the first time, dynamically determines where to begin the search in an ANN index on the fly, therefore exploiting query locality. CatapultDB injects shortcut edges called catapults that connect query regions to frequently visited destination nodes. Catapults are maintained as an additional layer on top of the graph, so the standard vector search algorithm remains unchanged: queries are simply routed to a better starting point when an appropriate catapult exists. This transparent design preserves the full feature set of the underlying system, including filtered search, dynamic insertions, and disk-resident indices. We implement CatapultDB and evaluate it using four workloads with varying amounts of bias. Our experiments show that CatapultDB increases throughput by up to 2.51x compared to DiskANN at equivalent or better recall, matches the efficiency of LSH-based approaches without sacrificing filtering or requiring index reconstruction, and adapts gracefully to workload shifts, unlike cache-based alternatives. △ Less

Submitted 2 March, 2026; originally announced March 2026.

Mosaic Learning: A Framework for Decentralized Learning with Model Fragmentation

Authors: Sayan Biswas, Davide Frey, Romaric Gaudel, Nirupam Gupta, Anne-Marie Kermarrec, Dimitri Lerévérend, Rafael Pires, Rishi Sharma, François Taïani, Martijn de Vos

Abstract: Decentralized learning (DL) enables collaborative machine learning (ML) without a central server, making it suitable for settings where training data cannot be centrally hosted. We introduce Mosaic Learning, a DL framework that decomposes models into fragments and disseminates them independently across the network. Fragmentation reduces redundant communication across correlated parameters and enab… ▽ More Decentralized learning (DL) enables collaborative machine learning (ML) without a central server, making it suitable for settings where training data cannot be centrally hosted. We introduce Mosaic Learning, a DL framework that decomposes models into fragments and disseminates them independently across the network. Fragmentation reduces redundant communication across correlated parameters and enables more diverse information propagation without increasing communication cost. We theoretically show that Mosaic Learning (i) shows state-of-the-art worst-case convergence rate, and (ii) leverages parameter correlation in an ML model, improving contraction by reducing the highest eigenvalue of a simplified system. We empirically evaluate Mosaic Learning on four learning tasks and observe up to 12 percentage points higher node-level test accuracy compared to epidemic learning (EL), a state-of-the-art baseline. In summary, Mosaic Learning improves DL performance without sacrificing its utility or efficiency, and positions itself as a new DL standard. △ Less

Submitted 4 February, 2026; originally announced February 2026.

Effective LoRA Adapter Routing using Task Representations

Authors: Akash Dhasade, Anne-Marie Kermarrec, Igor Pavlovic, Diana Petrescu, Rafael Pires, Mathis Randl, Martijn de Vos

Abstract: Low-rank adaptation (LoRA) enables parameter efficient specialization of large language models (LLMs) through modular adapters, resulting in rapidly growing public adapter pools spanning diverse tasks. Effectively using these adapters requires routing: selecting and composing the appropriate adapters for a query. We introduce LORAUTER, a novel routing framework that selects and composes LoRA adapt… ▽ More Low-rank adaptation (LoRA) enables parameter efficient specialization of large language models (LLMs) through modular adapters, resulting in rapidly growing public adapter pools spanning diverse tasks. Effectively using these adapters requires routing: selecting and composing the appropriate adapters for a query. We introduce LORAUTER, a novel routing framework that selects and composes LoRA adapters using task representations rather than adapter characteristics. Unlike existing approaches that map queries directly to adapters, LORAUTER routes queries via task embeddings derived from small validation sets and does not require adapter training data. By operating at the task level, LORAUTER achieves efficient routing that scales with the number of tasks rather than the number of adapters. Experiments across multiple tasks show that LORAUTER consistently outperforms baseline routing approaches, matching Oracle performance (101.2%) when task-aligned adapters exist and achieving state-of-the-art results on unseen tasks (+5.2 points). We further demonstrate the robustness of LORAUTER to very large, noisy adapter pools by scaling it to over 1500 adapters. △ Less

Submitted 30 January, 2026; v1 submitted 29 January, 2026; originally announced January 2026.

PoETa v2: Toward More Robust Evaluation of Large Language Models in Portuguese

Authors: Thales Sales Almeida, Ramon Pires, Hugo Abonizio, Rodrigo Nogueira, Hélio Pedrini

Abstract: Large Language Models (LLMs) exhibit significant variations in performance across linguistic and cultural contexts, underscoring the need for systematic evaluation in diverse languages. In this work, we present the most extensive evaluation of LLMs for the Portuguese language to date. Leveraging our newly introduced PoETa v2 benchmark -- a comprehensive suite of over 40 tasks in Portuguese -- we a… ▽ More Large Language Models (LLMs) exhibit significant variations in performance across linguistic and cultural contexts, underscoring the need for systematic evaluation in diverse languages. In this work, we present the most extensive evaluation of LLMs for the Portuguese language to date. Leveraging our newly introduced PoETa v2 benchmark -- a comprehensive suite of over 40 tasks in Portuguese -- we assess more than 20 models covering a broad spectrum of training scales and computational resources. Our study reveals how computational investment and language-specific adaptation impact performance in Portuguese, while also analyzing performance gaps in comparison to equivalent tasks in English. Through this benchmark and analysis, PoETa v2 lays the groundwork for future research on Portuguese language modeling and evaluation. The benchmark is available at https://github.com/PoETaV2/PoETaV2. △ Less

Submitted 26 November, 2025; v1 submitted 21 November, 2025; originally announced November 2025.

Efficient Pyramidal Analysis of Gigapixel Images on a Decentralized Modest Computer Cluster

Authors: Marie Reinbigler, Rishi Sharma, Rafael Pires, Elisabeth Brunet, Anne-Marie Kermarrec, Catalin Fetita

Abstract: Analyzing gigapixel images is recognized as computationally demanding. In this paper, we introduce PyramidAI, a technique for analyzing gigapixel images with reduced computational cost. The proposed approach adopts a gradual analysis of the image, beginning with lower resolutions and progressively concentrating on regions of interest for detailed examination at higher resolutions. We investigated… ▽ More Analyzing gigapixel images is recognized as computationally demanding. In this paper, we introduce PyramidAI, a technique for analyzing gigapixel images with reduced computational cost. The proposed approach adopts a gradual analysis of the image, beginning with lower resolutions and progressively concentrating on regions of interest for detailed examination at higher resolutions. We investigated two strategies for tuning the accuracy-computation performance trade-off when implementing the adaptive resolution selection, validated against the Camelyon16 dataset of biomedical images. Our results demonstrate that PyramidAI substantially decreases the amount of processed data required for analysis by up to 2.65x, while preserving the accuracy in identifying relevant sections on a single computer. To ensure democratization of gigapixel image analysis, we evaluated the potential to use mainstream computers to perform the computation by exploiting the parallelism potential of the approach. Using a simulator, we estimated the best data distribution and load balancing algorithm according to the number of workers. The selected algorithms were implemented and highlighted the same conclusions in a real-world setting. Analysis time is reduced from more than an hour to a few minutes using 12 modest workers, offering a practical solution for efficient large-scale image analysis. △ Less

Submitted 2 September, 2025; originally announced September 2025.

Comments: Accepted at the 31st International European Conference on Parallel and Distributed Computing (Euro-Par'25)

Exploitation Over Exploration: Unmasking the Bias in Linear Bandit Recommender Offline Evaluation

Authors: Pedro R. Pires, Gregorio F. Azevedo, Pietro L. Campos, Rafael T. Sereicikas, Tiago A. Almeida

Abstract: Multi-Armed Bandit (MAB) algorithms are widely used in recommender systems that require continuous, incremental learning. A core aspect of MABs is the exploration-exploitation trade-off: choosing between exploiting items likely to be enjoyed and exploring new ones to gather information. In contextual linear bandits, this trade-off is particularly central, as many variants share the same linear reg… ▽ More Multi-Armed Bandit (MAB) algorithms are widely used in recommender systems that require continuous, incremental learning. A core aspect of MABs is the exploration-exploitation trade-off: choosing between exploiting items likely to be enjoyed and exploring new ones to gather information. In contextual linear bandits, this trade-off is particularly central, as many variants share the same linear regression backbone and differ primarily in their exploration strategies. Despite its prevalent use, offline evaluation of MABs is increasingly recognized for its limitations in reliably assessing exploration behavior. This study conducts an extensive offline empirical comparison of several linear MABs. Strikingly, across over 90% of various datasets, a greedy linear model, with no type of exploration, consistently achieves top-tier performance, often outperforming or matching its exploratory counterparts. This observation is further corroborated by hyperparameter optimization, which consistently favors configurations that minimize exploration, suggesting that pure exploitation is the dominant strategy within these evaluation settings. Our results expose significant inadequacies in offline evaluation protocols for bandits, particularly concerning their capacity to reflect true exploratory efficacy. Consequently, this research underscores the urgent necessity for developing more robust assessment methodologies, guiding future investigations into alternative evaluation frameworks for interactive learning in recommender systems. △ Less

Submitted 16 April, 2026; v1 submitted 24 July, 2025; originally announced July 2025.

Comments: Published in RecSys'25, 10 pages, 3 figures

Interact2Vec -- An efficient neural network-based model for simultaneously learning users and items embeddings in recommender systems

Authors: Pedro R. Pires, Tiago A. Almeida

Abstract: Over the past decade, recommender systems have experienced a surge in popularity. Despite notable progress, they grapple with challenging issues, such as high data dimensionality and sparseness. Representing users and items as low-dimensional embeddings learned via neural networks has become a leading solution. However, while recent studies show promising results, many approaches rely on complex a… ▽ More Over the past decade, recommender systems have experienced a surge in popularity. Despite notable progress, they grapple with challenging issues, such as high data dimensionality and sparseness. Representing users and items as low-dimensional embeddings learned via neural networks has become a leading solution. However, while recent studies show promising results, many approaches rely on complex architectures or require content data, which may not always be available. This paper presents Interact2Vec, a novel neural network-based model that simultaneously learns distributed embeddings for users and items while demanding only implicit feedback. The model employs state-of-the-art strategies that natural language processing models commonly use to optimize the training phase and enhance the final embeddings. Two types of experiments were conducted regarding the extrinsic and intrinsic quality of the model. In the former, we benchmarked the recommendations generated by Interact2Vec's embeddings in a top-$N$ ranking problem, comparing them with six other recommender algorithms. The model achieved the second or third-best results in 30% of the datasets, being competitive with other recommenders, and has proven to be very efficient with an average training time reduction of 274% compared to other embedding-based models. Later, we analyzed the intrinsic quality of the embeddings through similarity tables. Our findings suggest that Interact2Vec can achieve promising results, especially on the extrinsic task, and is an excellent embedding-generator model for scenarios of scarce computing resources, enabling the learning of item and user embeddings simultaneously and efficiently. △ Less

Submitted 24 July, 2025; v1 submitted 27 June, 2025; originally announced June 2025.

Comments: Published in Applied Soft Computing (ASOC), 49 pages, 14 figures

Journal ref: Applied Soft Computing 181 (2025) 113408:1-113408:17

HarMoEny: Efficient Multi-GPU Inference of MoE Models

Authors: Zachary Doucet, Rishi Sharma, Martijn de Vos, Rafael Pires, Anne-Marie Kermarrec, Oana Balmau

Abstract: Mixture-of-Experts (MoE) models offer computational efficiency during inference by activating only a subset of specialized experts for a given input. This enables efficient model scaling on multi-GPU systems that use expert parallelism without compromising performance. However, load imbalance among experts and GPUs introduces waiting times, which can significantly increase inference latency. To ad… ▽ More Mixture-of-Experts (MoE) models offer computational efficiency during inference by activating only a subset of specialized experts for a given input. This enables efficient model scaling on multi-GPU systems that use expert parallelism without compromising performance. However, load imbalance among experts and GPUs introduces waiting times, which can significantly increase inference latency. To address this challenge, we propose HarMoEny, a novel solution to address MoE load imbalance through two simple techniques: (i) dynamic token redistribution to underutilized GPUs and (ii) asynchronous prefetching of experts from the system to GPU memory. These techniques achieve a near-perfect load balance among experts and GPUs and mitigate delays caused by overloaded GPUs. We implement HarMoEny and compare its latency and throughput with four MoE baselines using real-world and synthetic datasets. Under heavy load imbalance, HarMoEny increases throughput by 37%-70% and reduces time-to-first-token by 34%-41%, compared to the next-best baseline. Moreover, our ablation study demonstrates that HarMoEny's scheduling policy reduces the GPU idling time by up to 84% compared to the baseline policies. △ Less

Submitted 17 June, 2025; v1 submitted 14 June, 2025; originally announced June 2025.

Automatic Legal Writing Evaluation of LLMs

Authors: Ramon Pires, Roseval Malaquias Junior, Rodrigo Nogueira

Abstract: Despite the recent advances in Large Language Models, benchmarks for evaluating legal writing remain scarce due to the inherent complexity of assessing open-ended responses in this domain. One of the key challenges in evaluating language models on domain-specific tasks is finding test datasets that are public, frequently updated, and contain comprehensive evaluation guidelines. The Brazilian Bar E… ▽ More Despite the recent advances in Large Language Models, benchmarks for evaluating legal writing remain scarce due to the inherent complexity of assessing open-ended responses in this domain. One of the key challenges in evaluating language models on domain-specific tasks is finding test datasets that are public, frequently updated, and contain comprehensive evaluation guidelines. The Brazilian Bar Examination meets these requirements. We introduce oab-bench, a benchmark comprising 105 questions across seven areas of law from recent editions of the exam. The benchmark includes comprehensive evaluation guidelines and reference materials used by human examiners to ensure consistent grading. We evaluate the performance of four LLMs on oab-bench, finding that Claude-3.5 Sonnet achieves the best results with an average score of 7.93 out of 10, passing all 21 exams. We also investigated whether LLMs can serve as reliable automated judges for evaluating legal writing. Our experiments show that frontier models like OpenAI's o1 achieve a strong correlation with human scores when evaluating approved exams, suggesting their potential as reliable automated evaluators despite the inherently subjective nature of legal writing assessment. The source code and the benchmark -- containing questions, evaluation guidelines, model-generated responses, and their respective automated evaluations -- are publicly available. △ Less

Submitted 29 April, 2025; originally announced April 2025.

Multispectral airborne laser scanning for tree species classification: a benchmark of machine learning and deep learning algorithms

Authors: Josef Taher, Eric Hyyppä, Matti Hyyppä, Klaara Salolahti, Xiaowei Yu, Leena Matikainen, Antero Kukko, Matti Lehtomäki, Harri Kaartinen, Sopitta Thurachen, Paula Litkey, Ville Luoma, Markus Holopainen, Gefei Kong, Hongchao Fan, Petri Rönnholm, Matti Vaaja, Antti Polvivaara, Samuli Junttila, Mikko Vastaranta, Stefano Puliti, Rasmus Astrup, Joel Kostensalo, Mari Myllymäki, Maksymilian Kulicki , et al. (24 additional authors not shown)

Abstract: Climate-smart and biodiversity-preserving forestry demands precise information on forest resources, extending to the individual tree level. Multispectral airborne laser scanning (ALS) has shown promise in automated point cloud processing, but challenges remain in leveraging deep learning techniques and identifying rare tree species in class-imbalanced datasets. This study addresses these gaps by c… ▽ More Climate-smart and biodiversity-preserving forestry demands precise information on forest resources, extending to the individual tree level. Multispectral airborne laser scanning (ALS) has shown promise in automated point cloud processing, but challenges remain in leveraging deep learning techniques and identifying rare tree species in class-imbalanced datasets. This study addresses these gaps by conducting a comprehensive benchmark of deep learning and traditional shallow machine learning methods for tree species classification. For the study, we collected high-density multispectral ALS data ($>1000$ $\mathrm{pts}/\mathrm{m}^2$) at three wavelengths using the FGI-developed HeliALS system, complemented by existing Optech Titan data (35 $\mathrm{pts}/\mathrm{m}^2$), to evaluate the species classification accuracy of various algorithms in a peri-urban study area located in southern Finland. We established a field reference dataset of 6326 segments across nine species using a newly developed browser-based crowdsourcing tool, which facilitated efficient data annotation. The ALS data, including a training dataset of 1065 segments, was shared with the scientific community to foster collaborative research and diverse algorithmic contributions. Based on 5261 test segments, our findings demonstrate that point-based deep learning methods, particularly a point transformer model, outperformed traditional machine learning and image-based deep learning approaches on high-density multispectral point clouds. For the high-density ALS dataset, a point transformer model provided the best performance reaching an overall (macro-average) accuracy of 87.9% (74.5%) with a training set of 1065 segments and 92.0% (85.1%) with a larger training set of 5000 segments. △ Less

Submitted 17 February, 2026; v1 submitted 19 April, 2025; originally announced April 2025.

Journal ref: ISPRS Journal of Photogrammetry and Remote Sensing, Volume 233, 2026, Pages 278-309

Accelerating MoE Model Inference with Expert Sharding

Authors: Oana Balmau, Anne-Marie Kermarrec, Rafael Pires, André Loureiro Espírito Santo, Martijn de Vos, Milos Vujasinovic

Abstract: Mixture of experts (MoE) models achieve state-of-the-art results in language modeling but suffer from inefficient hardware utilization due to imbalanced token routing and communication overhead. While prior work has focused on optimizing MoE training and decoder architectures, inference for encoder-based MoE models in a multi-GPU with expert parallelism setting remains underexplored. We introduce… ▽ More Mixture of experts (MoE) models achieve state-of-the-art results in language modeling but suffer from inefficient hardware utilization due to imbalanced token routing and communication overhead. While prior work has focused on optimizing MoE training and decoder architectures, inference for encoder-based MoE models in a multi-GPU with expert parallelism setting remains underexplored. We introduce MoEShard, an inference system that achieves perfect load balancing through tensor sharding of MoE experts. Unlike existing approaches that rely on heuristic capacity factors or drop tokens, MoEShard evenly distributes computation across GPUs and ensures full token retention, maximizing utilization regardless of routing skewness. We achieve this through a strategic row- and column-wise decomposition of expert matrices. This reduces idle time and avoids bottlenecks caused by imbalanced expert assignments. Furthermore, MoEShard minimizes kernel launches by fusing decomposed expert computations, significantly improving throughput. We evaluate MoEShard against DeepSpeed on encoder-based architectures, demonstrating speedups of up to 6.4$\times$ in time to first token (TTFT). Our results show that tensor sharding, when properly applied to experts, is a viable and effective strategy for efficient MoE inference. △ Less

Submitted 11 March, 2025; originally announced March 2025.

Comments: To appear in the proceedings of the 5th Workshop on Machine Learning and Systems (EuroMLSys 25)

Leveraging Approximate Caching for Faster Retrieval-Augmented Generation

Authors: Shai Bergman, Anne-Marie Kermarrec, Diana Petrescu, Rafael Pires, Mathis Randl, Martijn de Vos, Ji Zhang

Abstract: Retrieval-augmented generation (RAG) improves the reliability of large language model (LLM) answers by integrating external knowledge. However, RAG increases the end-to-end inference time since looking for relevant documents from large vector databases is computationally expensive. To address this, we introduce Proximity, an approximate key-value cache that optimizes the RAG workflow by leveraging… ▽ More Retrieval-augmented generation (RAG) improves the reliability of large language model (LLM) answers by integrating external knowledge. However, RAG increases the end-to-end inference time since looking for relevant documents from large vector databases is computationally expensive. To address this, we introduce Proximity, an approximate key-value cache that optimizes the RAG workflow by leveraging similarities in user queries. Instead of treating each query independently, Proximity reuses previously retrieved documents when similar queries appear, substantially reducing the reliance on expensive vector database lookups. To efficiently scale, Proximity employs a locality-sensitive hashing (LSH) scheme that enables fast cache lookups while preserving retrieval accuracy. We evaluate Proximity using the MMLU and MedRAG question-answering benchmarks. Our experiments demonstrate that Proximity with our LSH scheme and a realistically-skewed MedRAG workload reduces database calls by 77.2% while maintaining database recall and test accuracy. We experiment with different similarity tolerances and cache capacities, and show that the time spent within the Proximity cache remains low and constant (4.8 microseconds) even as the cache grows substantially in size. Our results demonstrate that approximate caching is a practical and effective strategy for optimizing RAG-based systems. △ Less

Submitted 27 October, 2025; v1 submitted 7 March, 2025; originally announced March 2025.

Comments: Accepted at Middleware '25

Efficient Federated Search for Retrieval-Augmented Generation using Lightweight Routing

Authors: Akash Dhasade, Rachid Guerraoui, Anne-Marie Kermarrec, Diana Petrescu, Rafael Pires, Mathis Randl, Martijn de Vos

Abstract: Large language models (LLMs) achieve remarkable performance across domains but remain prone to hallucinations and inconsistencies. Retrieval-augmented generation (RAG) mitigates these issues by augmenting model inputs with relevant documents retrieved from external sources. In many real-world scenarios, relevant knowledge is fragmented across organizations or institutions, motivating the need for… ▽ More Large language models (LLMs) achieve remarkable performance across domains but remain prone to hallucinations and inconsistencies. Retrieval-augmented generation (RAG) mitigates these issues by augmenting model inputs with relevant documents retrieved from external sources. In many real-world scenarios, relevant knowledge is fragmented across organizations or institutions, motivating the need for federated search mechanisms that can aggregate results from heterogeneous data sources without centralizing the data. We introduce RAGRoute, a lightweight routing mechanism for federated search in RAG systems that dynamically selects relevant data sources at query time using a neural classifier, avoiding indiscriminate querying. This selective routing reduces communication overhead and end-to-end latency while preserving retrieval quality, achieving up to 80.65% reductions in communication volume and 52.50% reductions in latency across three benchmarks, while matching the accuracy of querying all sources. △ Less

Submitted 9 April, 2026; v1 submitted 26 February, 2025; originally announced February 2025.

Comments: To appear in the proceedings of DAIS 2026 (Distributed Applications and Interoperable Systems). An earlier version appeared at EuroMLSys 2025

The interplay between domain specialization and model size

Authors: Roseval Malaquias Junior, Ramon Pires, Thales Sales Almeida, Kenzo Sakiyama, Roseli A. F. Romero, Rodrigo Nogueira

Abstract: Scaling laws for language models have often focused on finding the optimal model size and token count for training from scratch. However, achieving this optimal balance requires significant compute resources due to the extensive data demands when training models from randomly-initialized weights. Continued pretraining offers a cost-effective alternative, leveraging the compute investment from pret… ▽ More Scaling laws for language models have often focused on finding the optimal model size and token count for training from scratch. However, achieving this optimal balance requires significant compute resources due to the extensive data demands when training models from randomly-initialized weights. Continued pretraining offers a cost-effective alternative, leveraging the compute investment from pretrained models to incorporate new knowledge without requiring extensive new data. Recent findings suggest that data quality influences constants in scaling laws, thereby altering the optimal parameter-token allocation ratio. Building on this insight, we investigate the interplay between domain specialization and model size during continued pretraining under compute-constrained scenarios. Our goal is to identify an optimal training regime for this scenario and detect patterns in this interplay that can be generalized across different model sizes and domains. To compare general and specialized training, we filtered a web-based dataset to extract data from three domains: legal, medical, and accounting. We pretrained models with 1.5B, 3B, 7B, and 14B parameters on both the unfiltered and filtered datasets, then evaluated their performance on domain-specific exams. Results show that as model size increases, specialized models outperform general models while requiring less training compute. Additionally, their growing compute efficiency leads to reduced forgetting of previously learned knowledge. △ Less

Submitted 29 March, 2025; v1 submitted 3 January, 2025; originally announced January 2025.

Revisiting Ensembling in One-Shot Federated Learning

Authors: Youssef Allouah, Akash Dhasade, Rachid Guerraoui, Nirupam Gupta, Anne-Marie Kermarrec, Rafael Pinot, Rafael Pires, Rishi Sharma

Abstract: Federated learning (FL) is an appealing approach to training machine learning models without sharing raw data. However, standard FL algorithms are iterative and thus induce a significant communication cost. One-shot federated learning (OFL) trades the iterative exchange of models between clients and the server with a single round of communication, thereby saving substantially on communication cost… ▽ More Federated learning (FL) is an appealing approach to training machine learning models without sharing raw data. However, standard FL algorithms are iterative and thus induce a significant communication cost. One-shot federated learning (OFL) trades the iterative exchange of models between clients and the server with a single round of communication, thereby saving substantially on communication costs. Not surprisingly, OFL exhibits a performance gap in terms of accuracy with respect to FL, especially under high data heterogeneity. We introduce FENS, a novel federated ensembling scheme that approaches the accuracy of FL with the communication efficiency of OFL. Learning in FENS proceeds in two phases: first, clients train models locally and send them to the server, similar to OFL; second, clients collaboratively train a lightweight prediction aggregator model using FL. We showcase the effectiveness of FENS through exhaustive experiments spanning several datasets and heterogeneity levels. In the particular case of heterogeneously distributed CIFAR-10 dataset, FENS achieves up to a 26.9% higher accuracy over state-of-the-art (SOTA) OFL, being only 3.1% lower than FL. At the same time, FENS incurs at most 4.3x more communication than OFL, whereas FL is at least 10.9x more communication-intensive than FENS. △ Less

Submitted 11 November, 2024; originally announced November 2024.

Comments: Accepted at NeurIPS 2024

Boosting Asynchronous Decentralized Learning with Model Fragmentation

Authors: Sayan Biswas, Anne-Marie Kermarrec, Alexis Marouani, Rafael Pires, Rishi Sharma, Martijn de Vos

Abstract: Decentralized learning (DL) is an emerging technique that allows nodes on the web to collaboratively train machine learning models without sharing raw data. Dealing with stragglers, i.e., nodes with slower compute or communication than others, is a key challenge in DL. We present DivShare, a novel asynchronous DL algorithm that achieves fast model convergence in the presence of communication strag… ▽ More Decentralized learning (DL) is an emerging technique that allows nodes on the web to collaboratively train machine learning models without sharing raw data. Dealing with stragglers, i.e., nodes with slower compute or communication than others, is a key challenge in DL. We present DivShare, a novel asynchronous DL algorithm that achieves fast model convergence in the presence of communication stragglers. DivShare achieves this by having nodes fragment their models into parameter subsets and send, in parallel to computation, each subset to a random sample of other nodes instead of sequentially exchanging full models. The transfer of smaller fragments allows more efficient usage of the collective bandwidth and enables nodes with slow network links to quickly contribute with at least some of their model parameters. By theoretically proving the convergence of DivShare, we provide, to the best of our knowledge, the first formal proof of convergence for a DL algorithm that accounts for the effects of asynchronous communication with delays. We experimentally evaluate DivShare against two state-of-the-art DL baselines, AD-PSGD and Swift, and with two standard datasets, CIFAR-10 and MovieLens. We find that DivShare with communication stragglers lowers time-to-accuracy by up to 3.9x compared to AD-PSGD on the CIFAR-10 dataset. Compared to baselines, DivShare also achieves up to 19.4% better accuracy and 9.5% lower test loss on the CIFAR-10 and MovieLens datasets, respectively. △ Less

Submitted 3 February, 2025; v1 submitted 16 October, 2024; originally announced October 2024.

Comments: Accepted to appear in the Proceedings of the ACM Web Conference 2025 (WWW '25)

Sabiá-3 Technical Report

Authors: Hugo Abonizio, Thales Sales Almeida, Thiago Laitz, Roseval Malaquias Junior, Giovana Kerche Bonás, Rodrigo Nogueira, Ramon Pires

Abstract: This report presents Sabiá-3, our new flagship language model, and Sabiazinho-3, a more cost-effective sibling. The models were trained on a large brazilian-centric corpus. Evaluations across diverse professional and academic benchmarks show a strong performance on Portuguese and Brazil-related tasks. Sabiá-3 shows large improvements in comparison to our previous best of model, Sabia-2 Medium, esp… ▽ More This report presents Sabiá-3, our new flagship language model, and Sabiazinho-3, a more cost-effective sibling. The models were trained on a large brazilian-centric corpus. Evaluations across diverse professional and academic benchmarks show a strong performance on Portuguese and Brazil-related tasks. Sabiá-3 shows large improvements in comparison to our previous best of model, Sabia-2 Medium, especially in reasoning-intensive tasks. Notably, Sabiá-3's average performance matches frontier LLMs, while it is offered at a three to four times lower cost per token, reinforcing the benefits of domain specialization. △ Less

Submitted 1 April, 2025; v1 submitted 15 October, 2024; originally announced October 2024.

Energy-Aware Decentralized Learning with Intermittent Model Training

Authors: Akash Dhasade, Paolo Dini, Elia Guerra, Anne-Marie Kermarrec, Marco Miozzo, Rafael Pires, Rishi Sharma, Martijn de Vos

Abstract: Decentralized learning (DL) offers a powerful framework where nodes collaboratively train models without sharing raw data and without the coordination of a central server. In the iterative rounds of DL, models are trained locally, shared with neighbors in the topology, and aggregated with other models received from neighbors. Sharing and merging models contribute to convergence towards a consensus… ▽ More Decentralized learning (DL) offers a powerful framework where nodes collaboratively train models without sharing raw data and without the coordination of a central server. In the iterative rounds of DL, models are trained locally, shared with neighbors in the topology, and aggregated with other models received from neighbors. Sharing and merging models contribute to convergence towards a consensus model that generalizes better across the collective data captured at training time. In addition, the energy consumption while sharing and merging model parameters is negligible compared to the energy spent during the training phase. Leveraging this fact, we present SkipTrain, a novel DL algorithm, which minimizes energy consumption in decentralized learning by strategically skipping some training rounds and substituting them with synchronization rounds. These training-silent periods, besides saving energy, also allow models to better mix and finally produce models with superior accuracy than typical DL algorithms that train at every round. Our empirical evaluations with 256 nodes demonstrate that SkipTrain reduces energy consumption by 50% and increases model accuracy by up to 12% compared to D-PSGD, the conventional DL algorithm. △ Less

Submitted 1 July, 2024; originally announced July 2024.

Harnessing Increased Client Participation with Cohort-Parallel Federated Learning

Authors: Akash Dhasade, Anne-Marie Kermarrec, Tuan-Anh Nguyen, Rafael Pires, Martijn de Vos

Abstract: Federated learning (FL) is a machine learning approach where nodes collaboratively train a global model. As more nodes participate in a round of FL, the effectiveness of individual model updates by nodes also diminishes. In this study, we increase the effectiveness of client updates by dividing the network into smaller partitions, or cohorts. We introduce Cohort-Parallel Federated Learning (CPFL):… ▽ More Federated learning (FL) is a machine learning approach where nodes collaboratively train a global model. As more nodes participate in a round of FL, the effectiveness of individual model updates by nodes also diminishes. In this study, we increase the effectiveness of client updates by dividing the network into smaller partitions, or cohorts. We introduce Cohort-Parallel Federated Learning (CPFL): a novel learning approach where each cohort independently trains a global model using FL, until convergence, and the produced models by each cohort are then unified using knowledge distillation. The insight behind CPFL is that smaller, isolated networks converge quicker than in a one-network setting where all nodes participate. Through exhaustive experiments involving realistic traces and non-IID data distributions on the CIFAR-10 and FEMNIST image classification tasks, we investigate the balance between the number of cohorts, model accuracy, training time, and compute resources. Compared to traditional FL, CPFL with four cohorts, non-IID data distribution, and CIFAR-10 yields a 1.9x reduction in train time and a 1.3x reduction in resource usage, with a minimal drop in test accuracy. △ Less

Submitted 10 March, 2025; v1 submitted 24 May, 2024; originally announced May 2024.

Comments: To appear in the proceedings of EuroMLSys'25

Communication-Efficient Secure Aggregation in Decentralized Learning

Authors: Sayan Biswas, Anne-Marie Kermarrec, Rafael Pires, Rishi Sharma, Milos Vujasinovic

Abstract: Decentralized learning (DL) enables participants to collaboratively train models without a central server, yet it faces significant scalability challenges that demand sparsification to reduce the prohibitive communication costs of peer-to-peer exchange. While secure aggregation effectively mitigates privacy risks in standard settings, it has remained fundamentally incompatible with sparsification… ▽ More Decentralized learning (DL) enables participants to collaboratively train models without a central server, yet it faces significant scalability challenges that demand sparsification to reduce the prohibitive communication costs of peer-to-peer exchange. While secure aggregation effectively mitigates privacy risks in standard settings, it has remained fundamentally incompatible with sparsification in decentralized networks due to the mismatch of indices across local updates, forcing a trade-off between communication efficiency and privacy. This paper introduces CESAR, a novel protocol that resolves this incompatibility by integrating secure aggregation and sparsification to provide provable defense against honest-but-curious and colluding adversaries. By coordinating masks over parameter intersections, CESAR supports node dropouts and robust privacy without central aggregation. Empirical evaluations on models up to 124 million parameters demonstrate that CESAR matches the accuracy of non-private baselines while cutting total data exchange by 66.7 % compared to a standard full-parameter decentralized protocol (D-PSGD). With TopK sparsification on IID data, CESAR even exceeds by 0.3 % the accuracy achieved by D-PSGD with sparsification. Collectively, these results establish CESAR as the first decentralized protocol to achieve both privacy and communication efficiency through secure aggregation in DL. △ Less

Submitted 31 July, 2026; v1 submitted 13 May, 2024; originally announced May 2024.

Comments: Extended version of a paper accepted at the 45th International Symposium on Reliable Distributed Systems (SRDS 2026)

Noiseless Privacy-Preserving Decentralized Learning

Authors: Sayan Biswas, Mathieu Even, Anne-Marie Kermarrec, Laurent Massoulie, Rafael Pires, Rishi Sharma, Martijn de Vos

Abstract: Decentralized learning (DL) enables collaborative learning without a server and without training data leaving the users' devices. However, the models shared in DL can still be used to infer training data. Conventional defenses such as differential privacy and secure aggregation fall short in effectively safeguarding user privacy in DL, either sacrificing model utility or efficiency. We introduce S… ▽ More Decentralized learning (DL) enables collaborative learning without a server and without training data leaving the users' devices. However, the models shared in DL can still be used to infer training data. Conventional defenses such as differential privacy and secure aggregation fall short in effectively safeguarding user privacy in DL, either sacrificing model utility or efficiency. We introduce Shatter, a novel DL approach in which nodes create virtual nodes (VNs) to disseminate chunks of their full model on their behalf. This enhances privacy by (i) preventing attackers from collecting full models from other nodes, and (ii) hiding the identity of the original node that produced a given model chunk. We theoretically prove the convergence of Shatter and provide a formal analysis demonstrating how Shatter reduces the efficacy of attacks compared to when exchanging full models between nodes. We evaluate the convergence and attack resilience of Shatter with existing DL algorithms, with heterogeneous datasets, and against three standard privacy attacks. Our evaluation shows that Shatter not only renders these privacy attacks infeasible when each node operates 16 VNs but also exhibits a positive impact on model utility compared to standard DL. In summary, Shatter enhances the privacy of DL while maintaining the utility and efficiency of the model. △ Less

Submitted 12 September, 2024; v1 submitted 15 April, 2024; originally announced April 2024.

Comments: Accepted at PETS 2025

Juru: Legal Brazilian Large Language Model from Reputable Sources

Authors: Roseval Malaquias Junior, Ramon Pires, Roseli Romero, Rodrigo Nogueira

Abstract: The high compute cost associated with pretraining large language models limits their research. Two strategies have emerged to address this issue: domain specialization and pretraining with high-quality data. To explore these strategies, we specialized the Mistral-7B model with 1.9 billion unique tokens from reputable Brazilian legal sources and conducted few-shot evaluations on legal and general k… ▽ More The high compute cost associated with pretraining large language models limits their research. Two strategies have emerged to address this issue: domain specialization and pretraining with high-quality data. To explore these strategies, we specialized the Mistral-7B model with 1.9 billion unique tokens from reputable Brazilian legal sources and conducted few-shot evaluations on legal and general knowledge test suites. Our model, Juru, demonstrates the benefits of domain specialization by achieving improved performance on legal benchmarks, even with a reduced amount of pretraining data. However, this domain specialization through continued pretraining comes at the cost of increased forgetting in unrelated domains, as evidenced by performance degradation on general knowledge test suites in both Portuguese and English. This study contributes to the growing body of scientific evidence showing that pretraining data selection may enhance the performance of large language models, enabling the exploration of these models at a lower cost. Juru is publicly available at https://huggingface.co/roseval/Juru-7B . △ Less

Submitted 27 July, 2025; v1 submitted 26 March, 2024; originally announced March 2024.

Low-Cost Privacy-Preserving Decentralized Learning

Authors: Sayan Biswas, Davide Frey, Romaric Gaudel, Anne-Marie Kermarrec, Dimitri Lerévérend, Rafael Pires, Rishi Sharma, François Taïani

Abstract: Decentralized learning (DL) is an emerging paradigm of collaborative machine learning that enables nodes in a network to train models collectively without sharing their raw data or relying on a central server. This paper introduces Zip-DL, a privacy-aware DL algorithm that leverages correlated noise to achieve robust privacy against local adversaries while ensuring efficient convergence at low com… ▽ More Decentralized learning (DL) is an emerging paradigm of collaborative machine learning that enables nodes in a network to train models collectively without sharing their raw data or relying on a central server. This paper introduces Zip-DL, a privacy-aware DL algorithm that leverages correlated noise to achieve robust privacy against local adversaries while ensuring efficient convergence at low communication costs. By progressively neutralizing the noise added during distributed averaging, Zip-DL combines strong privacy guarantees with high model accuracy. Its design requires only one communication round per gradient descent iteration, significantly reducing communication overhead compared to competitors. We establish theoretical bounds on both convergence speed and privacy guarantees. Moreover, extensive experiments demonstrating Zip-DL's practical applicability make it outperform state-of-the-art methods in the accuracy vs. vulnerability trade-off. Specifically, Zip-DL (i) reduces membership-inference attack success rates by up to 35% compared to baseline DL, (ii) decreases attack efficacy by up to 13% compared to competitors offering similar utility, and (iii) achieves up to 59% higher accuracy to completely nullify a basic attack scenario, compared to a state-of-the-art privacy-preserving approach under the same threat model. These results position Zip-DL as a practical and efficient solution for privacy-preserving decentralized learning in real-world applications. △ Less

Submitted 11 March, 2025; v1 submitted 18 March, 2024; originally announced March 2024.

Comments: 24 pages, accepted at Pets 2025

Sabiá-2: A New Generation of Portuguese Large Language Models

Authors: Thales Sales Almeida, Hugo Abonizio, Rodrigo Nogueira, Ramon Pires

Abstract: We introduce Sabiá-2, a family of large language models trained on Portuguese texts. The models are evaluated on a diverse range of exams, including entry-level tests for Brazilian universities, professional certification exams, and graduate-level exams for various disciplines such as accounting, economics, engineering, law and medicine. Our results reveal that our best model so far, Sabiá-2 Mediu… ▽ More We introduce Sabiá-2, a family of large language models trained on Portuguese texts. The models are evaluated on a diverse range of exams, including entry-level tests for Brazilian universities, professional certification exams, and graduate-level exams for various disciplines such as accounting, economics, engineering, law and medicine. Our results reveal that our best model so far, Sabiá-2 Medium, matches or surpasses GPT-4's performance in 23 out of 64 exams and outperforms GPT-3.5 in 58 out of 64 exams. Notably, specialization has a significant impact on a model's performance without the need to increase its size, allowing us to offer Sabiá-2 Medium at a price per token that is 10 times cheaper than GPT-4. Finally, we identified that math and coding are key abilities that need improvement. △ Less

Submitted 26 March, 2024; v1 submitted 14 March, 2024; originally announced March 2024.

Connectedness in weighted consensus division of graphical cakes between two agents

Authors: Josef Hanke, Ana Rita Pires

Abstract: Austin's moving knife procedure was originally introduced to find a consensus division of an interval/circular cake between two agents, each of whom believes that they receive exactly half of the cake. We generalise this in two ways: we consider cakes modelled by graphs, and let the two agents have unequal, arbitrary entitlements. In this setting, we seek a weighted consensus division -- one where… ▽ More Austin's moving knife procedure was originally introduced to find a consensus division of an interval/circular cake between two agents, each of whom believes that they receive exactly half of the cake. We generalise this in two ways: we consider cakes modelled by graphs, and let the two agents have unequal, arbitrary entitlements. In this setting, we seek a weighted consensus division -- one where each agent believes they received exactly the share they are entitled to -- which also minimises the number of connected components that each agent receives. First, we review the weighted consensus division of a circular cake, which gives exactly one connected piece to each agent. Next, by judiciously mapping a circle to a graph, we produce a weighted consensus division of a star graph cake that gives at most two connected pieces to each agent -- and show that this bound on the number of connected pieces is tight. For a tree, each agent receives at most h+1 connected pieces, where h is the minimal height of the tree. For a connected graphical cake, each agent receives r+2 connected pieces, where r is the radius of the graph. Finally, for a graphical cake with s connected components, the division involves at most s+2r+4 connected pieces, where r is the maximum radius among all connected components. △ Less

Submitted 25 September, 2025; v1 submitted 19 December, 2023; originally announced December 2023.

Comments: This version agrees with the published version. 11 pages, 4 figures

Journal ref: Mathematical Social Sciences Mathematical Social Sciences, Volume 138, December 2025, 102463

Evaluating GPT-4's Vision Capabilities on Brazilian University Admission Exams

Authors: Ramon Pires, Thales Sales Almeida, Hugo Abonizio, Rodrigo Nogueira

Abstract: Recent advancements in language models have showcased human-comparable performance in academic entrance exams. However, existing studies often overlook questions that require the integration of visual comprehension, thus compromising the full spectrum and complexity inherent in real-world scenarios. To address this gap, we present a comprehensive framework to evaluate language models on entrance e… ▽ More Recent advancements in language models have showcased human-comparable performance in academic entrance exams. However, existing studies often overlook questions that require the integration of visual comprehension, thus compromising the full spectrum and complexity inherent in real-world scenarios. To address this gap, we present a comprehensive framework to evaluate language models on entrance exams, which incorporates both textual and visual elements. We evaluate the two most recent editions of Exame Nacional do Ensino Médio (ENEM), the main standardized entrance examination adopted by Brazilian universities. Our study not only reaffirms the capabilities of GPT-4 as the state of the art for handling complex multidisciplinary questions, but also pioneers in offering a realistic assessment of multimodal language models on Portuguese examinations. One of the highlights is that text captions transcribing visual content outperform the direct use of images, suggesting that the vision model has room for improvement. Yet, despite improvements afforded by images or captions, mathematical questions remain a challenge for these state-of-the-art models. The code and data used on experiments are available at https://github.com/piresramon/gpt-4-enem. △ Less

Submitted 23 November, 2023; originally announced November 2023.

Comments: arXiv admin note: substantial text overlap with arXiv:2303.17003

Epidemic Learning: Boosting Decentralized Learning with Randomized Communication

Authors: Martijn de Vos, Sadegh Farhadkhani, Rachid Guerraoui, Anne-Marie Kermarrec, Rafael Pires, Rishi Sharma

Abstract: We present Epidemic Learning (EL), a simple yet powerful decentralized learning (DL) algorithm that leverages changing communication topologies to achieve faster model convergence compared to conventional DL approaches. At each round of EL, each node sends its model updates to a random sample of $s$ other nodes (in a system of $n$ nodes). We provide an extensive theoretical analysis of EL, demonst… ▽ More We present Epidemic Learning (EL), a simple yet powerful decentralized learning (DL) algorithm that leverages changing communication topologies to achieve faster model convergence compared to conventional DL approaches. At each round of EL, each node sends its model updates to a random sample of $s$ other nodes (in a system of $n$ nodes). We provide an extensive theoretical analysis of EL, demonstrating that its changing topology culminates in superior convergence properties compared to the state-of-the-art (static and dynamic) topologies. Considering smooth non-convex loss functions, the number of transient iterations for EL, i.e., the rounds required to achieve asymptotic linear speedup, is in $O(n^3/s^2)$ which outperforms the best-known bound $O(n^3)$ by a factor of $s^2$, indicating the benefit of randomized communication for DL. We empirically evaluate EL in a 96-node network and compare its performance with state-of-the-art DL approaches. Our results illustrate that EL converges up to $ 1.7\times$ quicker than baseline DL algorithms and attains $2.2 $\% higher accuracy for the same communication volume. △ Less

Submitted 27 October, 2023; v1 submitted 3 October, 2023; originally announced October 2023.

Comments: Accepted paper at NeurIPS 2023

Get More for Less in Decentralized Learning Systems

Authors: Akash Dhasade, Anne-Marie Kermarrec, Rafael Pires, Rishi Sharma, Milos Vujasinovic, Jeffrey Wigger

Abstract: Decentralized learning (DL) systems have been gaining popularity because they avoid raw data sharing by communicating only model parameters, hence preserving data confidentiality. However, the large size of deep neural networks poses a significant challenge for decentralized training, since each node needs to exchange gigabytes of data, overloading the network. In this paper, we address this chall… ▽ More Decentralized learning (DL) systems have been gaining popularity because they avoid raw data sharing by communicating only model parameters, hence preserving data confidentiality. However, the large size of deep neural networks poses a significant challenge for decentralized training, since each node needs to exchange gigabytes of data, overloading the network. In this paper, we address this challenge with JWINS, a communication-efficient and fully decentralized learning system that shares only a subset of parameters through sparsification. JWINS uses wavelet transform to limit the information loss due to sparsification and a randomized communication cut-off that reduces communication usage without damaging the performance of trained models. We demonstrate empirically with 96 DL nodes on non-IID datasets that JWINS can achieve similar accuracies to full-sharing DL while sending up to 64% fewer bytes. Additionally, on low communication budgets, JWINS outperforms the state-of-the-art communication-efficient DL algorithm CHOCO-SGD by up to 4x in terms of network savings and time. △ Less

Submitted 31 October, 2023; v1 submitted 7 June, 2023; originally announced June 2023.

Journal ref: 2023 IEEE 43rd International Conference on Distributed Computing Systems (ICDCS 2023)

Decentralized Learning Made Easy with DecentralizePy

Authors: Akash Dhasade, Anne-Marie Kermarrec, Rafael Pires, Rishi Sharma, Milos Vujasinovic

Abstract: Decentralized learning (DL) has gained prominence for its potential benefits in terms of scalability, privacy, and fault tolerance. It consists of many nodes that coordinate without a central server and exchange millions of parameters in the inherently iterative process of machine learning (ML) training. In addition, these nodes are connected in complex and potentially dynamic topologies. Assessin… ▽ More Decentralized learning (DL) has gained prominence for its potential benefits in terms of scalability, privacy, and fault tolerance. It consists of many nodes that coordinate without a central server and exchange millions of parameters in the inherently iterative process of machine learning (ML) training. In addition, these nodes are connected in complex and potentially dynamic topologies. Assessing the intricate dynamics of such networks is clearly not an easy task. Often in literature, researchers resort to simulated environments that do not scale and fail to capture practical and crucial behaviors, including the ones associated to parallelism, data transfer, network delays, and wall-clock time. In this paper, we propose DecentralizePy, a distributed framework for decentralized ML, which allows for the emulation of large-scale learning networks in arbitrary topologies. We demonstrate the capabilities of DecentralizePy by deploying techniques such as sparsification and secure aggregation on top of several topologies, including dynamic networks with more than one thousand nodes. △ Less

Submitted 17 April, 2023; originally announced April 2023.

Journal ref: 2023 3rd European Workshop on Machine Learning and Systems (EuroMLSys 2023)

Sabiá: Portuguese Large Language Models

Authors: Ramon Pires, Hugo Abonizio, Thales Sales Almeida, Rodrigo Nogueira

Abstract: As the capabilities of language models continue to advance, it is conceivable that "one-size-fits-all" model will remain as the main paradigm. For instance, given the vast number of languages worldwide, many of which are low-resource, the prevalent practice is to pretrain a single model on multiple languages. In this paper, we add to the growing body of evidence that challenges this practice, demo… ▽ More As the capabilities of language models continue to advance, it is conceivable that "one-size-fits-all" model will remain as the main paradigm. For instance, given the vast number of languages worldwide, many of which are low-resource, the prevalent practice is to pretrain a single model on multiple languages. In this paper, we add to the growing body of evidence that challenges this practice, demonstrating that monolingual pretraining on the target language significantly improves models already extensively trained on diverse corpora. More specifically, we further pretrain GPT-J and LLaMA models on Portuguese texts using 3% or less of their original pretraining budget. Few-shot evaluations on Poeta, a suite of 14 Portuguese datasets, reveal that our models outperform English-centric and multilingual counterparts by a significant margin. Our best model, Sabiá-65B, performs on par with GPT-3.5-turbo. By evaluating on datasets originally conceived in the target language as well as translated ones, we study the contributions of language-specific pretraining in terms of 1) capturing linguistic nuances and structures inherent to the target language, and 2) enriching the model's knowledge about a domain or culture. Our results indicate that the majority of the benefits stem from the domain-specific knowledge acquired through monolingual pretraining. △ Less

Submitted 9 November, 2023; v1 submitted 16 April, 2023; originally announced April 2023.

Evaluating GPT-3.5 and GPT-4 Models on Brazilian University Admission Exams

Authors: Desnes Nunes, Ricardo Primi, Ramon Pires, Roberto Lotufo, Rodrigo Nogueira

Abstract: The present study aims to explore the capabilities of Language Models (LMs) in tackling high-stakes multiple-choice tests, represented here by the Exame Nacional do Ensino Médio (ENEM), a multidisciplinary entrance examination widely adopted by Brazilian universities. This exam poses challenging tasks for LMs, since its questions may span into multiple fields of knowledge, requiring understanding… ▽ More The present study aims to explore the capabilities of Language Models (LMs) in tackling high-stakes multiple-choice tests, represented here by the Exame Nacional do Ensino Médio (ENEM), a multidisciplinary entrance examination widely adopted by Brazilian universities. This exam poses challenging tasks for LMs, since its questions may span into multiple fields of knowledge, requiring understanding of information from diverse domains. For instance, a question may require comprehension of both statistics and biology to be solved. This work analyzed responses generated by GPT-3.5 and GPT-4 models for questions presented in the 2009-2017 exams, as well as for questions of the 2022 exam, which were made public after the training of the models was completed. Furthermore, different prompt strategies were tested, including the use of Chain-of-Thought (CoT) prompts to generate explanations for answers. On the 2022 edition, the best-performing model, GPT-4 with CoT, achieved an accuracy of 87%, largely surpassing GPT-3.5 by 11 points. The code and data used on experiments are available at https://github.com/piresramon/gpt-4-enem. △ Less

Submitted 29 March, 2023; originally announced March 2023.

Video Segmentation Learning Using Cascade Residual Convolutional Neural Network

Authors: Daniel F. S. Santos, Rafael G. Pires, Danilo Colombo, João P. Papa

Abstract: Video segmentation consists of a frame-by-frame selection process of meaningful areas related to foreground moving objects. Some applications include traffic monitoring, human tracking, action recognition, efficient video surveillance, and anomaly detection. In these applications, it is not rare to face challenges such as abrupt changes in weather conditions, illumination issues, shadows, subtle d… ▽ More Video segmentation consists of a frame-by-frame selection process of meaningful areas related to foreground moving objects. Some applications include traffic monitoring, human tracking, action recognition, efficient video surveillance, and anomaly detection. In these applications, it is not rare to face challenges such as abrupt changes in weather conditions, illumination issues, shadows, subtle dynamic background motions, and also camouflage effects. In this work, we address such shortcomings by proposing a novel deep learning video segmentation approach that incorporates residual information into the foreground detection learning process. The main goal is to provide a method capable of generating an accurate foreground detection given a grayscale video. Experiments conducted on the Change Detection 2014 and on the private dataset PetrobrasROUTES from Petrobras support the effectiveness of the proposed approach concerning some state-of-the-art video segmentation techniques, with overall F-measures of $\mathbf{0.9535}$ and $\mathbf{0.9636}$ in the Change Detection 2014 and PetrobrasROUTES datasets, respectively. Such a result places the proposed technique amongst the top 3 state-of-the-art video segmentation methods, besides comprising approximately seven times less parameters than its top one counterpart. △ Less

Submitted 20 December, 2022; originally announced December 2022.

Comments: Published in: 2019 32nd SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI). arXiv admin note: text overlap with arXiv:2212.10417

Scene Change Detection Using Multiscale Cascade Residual Convolutional Neural Networks

Authors: Daniel F. S. Santos, Rafael G. Pires, Danilo Colombo, João P. Papa

Abstract: Scene change detection is an image processing problem related to partitioning pixels of a digital image into foreground and background regions. Mostly, visual knowledge-based computer intelligent systems, like traffic monitoring, video surveillance, and anomaly detection, need to use change detection techniques. Amongst the most prominent detection methods, there are the learning-based ones, which… ▽ More Scene change detection is an image processing problem related to partitioning pixels of a digital image into foreground and background regions. Mostly, visual knowledge-based computer intelligent systems, like traffic monitoring, video surveillance, and anomaly detection, need to use change detection techniques. Amongst the most prominent detection methods, there are the learning-based ones, which besides sharing similar training and testing protocols, differ from each other in terms of their architecture design strategies. Such architecture design directly impacts on the quality of the detection results, and also in the device resources capacity, like memory. In this work, we propose a novel Multiscale Cascade Residual Convolutional Neural Network that integrates multiscale processing strategy through a Residual Processing Module, with a Segmentation Convolutional Neural Network. Experiments conducted on two different datasets support the effectiveness of the proposed approach, achieving average overall $\boldsymbol{F\text{-}measure}$ results of $\boldsymbol{0.9622}$ and $\boldsymbol{0.9664}$ over Change Detection 2014 and PetrobrasROUTES datasets respectively, besides comprising approximately eight times fewer parameters. Such obtained results place the proposed technique amongst the top four state-of-the-art scene change detection methods. △ Less

Submitted 20 December, 2022; originally announced December 2022.

Comments: Published in: 2020 33rd SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI)

DDIPNet and DDIPNet+: Discriminant Deep Image Prior Networks for Remote Sensing Image Classification

Authors: Daniel F. S. Santos, Rafael G. Pires, Leandro A. Passos, João P. Papa

Abstract: Research on remote sensing image classification significantly impacts essential human routine tasks such as urban planning and agriculture. Nowadays, the rapid advance in technology and the availability of many high-quality remote sensing images create a demand for reliable automation methods. The current paper proposes two novel deep learning-based architectures for image classification purposes,… ▽ More Research on remote sensing image classification significantly impacts essential human routine tasks such as urban planning and agriculture. Nowadays, the rapid advance in technology and the availability of many high-quality remote sensing images create a demand for reliable automation methods. The current paper proposes two novel deep learning-based architectures for image classification purposes, i.e., the Discriminant Deep Image Prior Network and the Discriminant Deep Image Prior Network+, which combine Deep Image Prior and Triplet Networks learning strategies. Experiments conducted over three well-known public remote sensing image datasets achieved state-of-the-art results, evidencing the effectiveness of using deep image priors for remote sensing image classification. △ Less

Submitted 20 December, 2022; originally announced December 2022.

Comments: Published in: 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS

TEE-based decentralized recommender systems: The raw data sharing redemption

Authors: Akash Dhasade, Nevena Dresevic, Anne-Marie Kermarrec, Rafael Pires

Abstract: Recommenders are central in many applications today. The most effective recommendation schemes, such as those based on collaborative filtering (CF), exploit similarities between user profiles to make recommendations, but potentially expose private data. Federated learning and decentralized learning systems address this by letting the data stay on user's machines to preserve privacy: each user perf… ▽ More Recommenders are central in many applications today. The most effective recommendation schemes, such as those based on collaborative filtering (CF), exploit similarities between user profiles to make recommendations, but potentially expose private data. Federated learning and decentralized learning systems address this by letting the data stay on user's machines to preserve privacy: each user performs the training on local data and only the model parameters are shared. However, sharing the model parameters across the network may still yield privacy breaches. In this paper, we present REX, the first enclave-based decentralized CF recommender. REX exploits Trusted execution environments (TEE), such as Intel software guard extensions (SGX), that provide shielded environments within the processor to improve convergence while preserving privacy. Firstly, REX enables raw data sharing, which ultimately speeds up convergence and reduces the network load. Secondly, REX fully preserves privacy. We analyze the impact of raw data sharing in both deep neural network (DNN) and matrix factorization (MF) recommenders and showcase the benefits of trusted environments in a full-fledged implementation of REX. Our experimental results demonstrate that through raw data sharing, REX significantly decreases the training time by 18.3x and the network load by 2 orders of magnitude over standard decentralized approaches that share only parameters, while fully protecting privacy by leveraging trustworthy hardware enclaves with very little overhead. △ Less

Submitted 1 June, 2022; v1 submitted 23 February, 2022; originally announced February 2022.

Journal ref: 2022 IEEE 36th International Parallel and Distributed Processing Symposium (IPDPS 2022) 447-458

Sequence-to-Sequence Models for Extracting Information from Registration and Legal Documents

Authors: Ramon Pires, Fábio C. de Souza, Guilherme Rosa, Roberto A. Lotufo, Rodrigo Nogueira

Abstract: A typical information extraction pipeline consists of token- or span-level classification models coupled with a series of pre- and post-processing scripts. In a production pipeline, requirements often change, with classes being added and removed, which leads to nontrivial modifications to the source code and the possible introduction of bugs. In this work, we evaluate sequence-to-sequence models a… ▽ More A typical information extraction pipeline consists of token- or span-level classification models coupled with a series of pre- and post-processing scripts. In a production pipeline, requirements often change, with classes being added and removed, which leads to nontrivial modifications to the source code and the possible introduction of bugs. In this work, we evaluate sequence-to-sequence models as an alternative to token-level classification methods for information extraction of legal and registration documents. We finetune models that jointly extract the information and generate the output already in a structured format. Post-processing steps are learned during training, thus eliminating the need for rule-based methods and simplifying the pipeline. Furthermore, we propose a novel method to align the output with the input text, thus facilitating system inspection and auditing. Our experiments on four real-world datasets show that the proposed method is an alternative to classical pipelines. △ Less

Submitted 14 January, 2022; originally announced January 2022.

Gait Recognition Based on Deep Learning: A Survey

Authors: Claudio Filipi Gonçalves dos Santos, Diego de Souza Oliveira, Leandro A. Passos, Rafael Gonçalves Pires, Daniel Felipe Silva Santos, Lucas Pascotti Valem, Thierry P. Moreira, Marcos Cleison S. Santana, Mateus Roder, João Paulo Papa, Danilo Colombo

Abstract: In general, biometry-based control systems may not rely on individual expected behavior or cooperation to operate appropriately. Instead, such systems should be aware of malicious procedures for unauthorized access attempts. Some works available in the literature suggest addressing the problem through gait recognition approaches. Such methods aim at identifying human beings through intrinsic perce… ▽ More In general, biometry-based control systems may not rely on individual expected behavior or cooperation to operate appropriately. Instead, such systems should be aware of malicious procedures for unauthorized access attempts. Some works available in the literature suggest addressing the problem through gait recognition approaches. Such methods aim at identifying human beings through intrinsic perceptible features, despite dressed clothes or accessories. Although the issue denotes a relatively long-time challenge, most of the techniques developed to handle the problem present several drawbacks related to feature extraction and low classification rates, among other issues. However, deep learning-based approaches recently emerged as a robust set of tools to deal with virtually any image and computer-vision related problem, providing paramount results for gait recognition as well. Therefore, this work provides a surveyed compilation of recent works regarding biometric detection through gait recognition with a focus on deep learning approaches, emphasizing their benefits, and exposing their weaknesses. Besides, it also presents categorized and characterized descriptions of the datasets, approaches, and architectures employed to tackle associated constraints. △ Less

Submitted 10 January, 2022; originally announced January 2022.