research
122 notes, newest first.
Inducing Task Models from Computer-Use Traces
Researchers present a new method called Task Model Induction (TMI) to extract detailed and structured task models from computer use traces, such as screenshots and mouse or keyboard actions. These task models can help…
MidTool: Mid-training Data Synthesis for Agentic Tool Use
In the paper 'MidTool: Mid-training Data Synthesis for Agentic Tool Use', researchers present a new open corpus construction pipeline called MidTool. This pipeline is designed to teach large language models how to use…
Metrics That Write Themselves: Evolving an Evaluator from Its Own Blind Spots
Researchers developed a method called EvalCEGAR to automatically create an evaluation metric for AI agents in tasks like report generation, where traditional metrics are difficult to develop. The method involves…
Self-prompting and cross-model consensus enable reproducible data extraction from scientific literature with large language models
The paper investigates the performance of large language models (LLMs) in extracting contextualized information from research articles, finding that while LLMs can perform well with expert-curated prompts, they may…
Browse articles
The article titled 'MultiMed-RAG: leveraging multi-source knowledge and agent collaboration for medical tasks' discusses a new system called CliniSense, which uses artificial intelligence to evaluate clinical skills…
ACL Anthology team
The ACL Anthology is a collection of papers related to natural language processing, published by the ACL Anthology team. Papers from 2016 onwards are licensed under Creative Commons Attribution 4.0 International…
Beyond Trial-and-Error: Agentic Optimization for Image-to-Video Adherence
Researchers present the 'Agentic Self-Improvement' framework to address challenges in professional workflows with modern Image-to-Video (I2V) models. The framework uses a two-stage approach: an iterative prompt…
Beyond Trial-and-Error: Agentic Optimization for Image-to-Video Adherence
Researchers have developed a new framework called 'Agentic Self-Improvement' to improve the performance of Image-to-Video (I2V) models, which are commonly used for automated content creation. The new framework aims to…
Diagram-MMU: A Multi-Modal Benchmark for Scientific Diagrams
A new benchmark called Diagram-MMU has been created to test the ability of Multimodal Large Language Models (MLLMs) in parsing and understanding scientific diagrams. The benchmark includes 3,700 curated diagrams and…
DreamFly: Causal Memory and Receding-Horizon Diffusion Planning for Aerial Vision-Language Navigation
Researchers propose DreamFly, a framework for aerial vision-language navigation that addresses challenges like limited historical context, short planning horizons, and unreliable implicit termination. The framework…
Preference Tree Optimization: Enhancing Goal-Oriented Dialogue with Look-Ahead Simulations
The research presents a new framework called Preference Tree Optimization (PTO) for enhancing dialogue systems in specialized domains with limited data. It focuses on Motivational Interviewing (MI), a counseling…
DuplexWorld: Can voice agents help you get through the day?
A research paper titled 'DuplexWorld' discusses the use of speech-to-speech (S2S) voice agents in various sectors like banking, insurance, travel, healthcare, logistics, and pathfinding. The study evaluates the…
IO Factory: Simulating AI-Enabled Influence Campaigns at Scale
The paper introduces IO Factory, an AI framework designed to simulate information and influence campaigns at scale. The framework is designed to handle not just individual language models but also AI swarms, which are…
Most biomedical publications show signs of LLM-assisted writing
A study suggests that by the end of 2025, 89% of biomedical papers on PubMed Central showed an excess of language associated with LLM-assisted writing. The authors argue that this could impact future guidelines and…
Quantum Coordination Advantages in AI State-Tracking Tasks: Semantic Compilation and Latent Memory
The paper presents a proof of quantum coordination advantages in AI state-tracking tasks, specifically in tasks involving semantic compilation and latent memory. It demonstrates that quantum computers can compress…
Beyond Naturalness: Probing Automated Text-To-Speech Evaluators on Linguistically Grounded Dimensions
Researchers create a new evaluation benchmark for Text-to-Speech (TTS) systems to measure their performance on specific linguistic dimensions beyond simple 'naturalness'. They find that current methods struggle to…
Hallucination-Free GUI Grounding via Regression-Free Layout-Aware Matching
Researchers propose a new framework for grounding user instructions in graphical user interfaces (GUI). The framework uses a frozen multimodal large language model (MLLM) to parse instructions and a dedicated grounding…
Open Evaluation Agent: Efficient and Promptable Evaluation of Visual Generative Models
Researchers present the Evaluation Agent framework, a system that efficiently evaluates visual generative models using human-like strategies. It decomposes evaluation requests into sub-aspects, generates prompts…
SHE: Trajectory-driven Safety Harness Evolution for LLM Agents
Researchers propose a framework called Safety Harness Evolution (SHE) to improve the safety of large language model agents. SHE learns evolving safe boundaries from rollout trajectories, allowing for localized harness…
Comparative Approaches to Agent Retrieval over Large Skill Libraries
The research compares two systems for deciding which skills to load from a large skill library when creating agents. One system combines lexical and dense-embedding retrieval, while the other uses a typed knowledge…
EnvACE: Internalizing Environment Dynamics via World Rehearsal for Agentic Reinforcement Learning
Researchers have introduced EnvACE, a method for training large language model agents using world rehearsal instead of interactions with real or simulated environments. The method involves the policy alternating…
Investigating Artificial Intelligence Digital Sovereignty in Mobile Shopping Apps: A Case Study of Nigeria
The research investigates how Artificial Intelligence (AI) in Nigerian mobile shopping apps affects digital sovereignty, focusing on platform transparency as a key indicator of user awareness and control. The findings…
QuanTiMedAI: Quantum-Enhanced Time-Series Model guided by Agentic AI for Cardiac Arrest Mortality Prediction
The researchers present QuanTiMedAI, a quantum-agentic framework designed for predicting cardiac arrest mortality rates. This system combines an agentic large language model for clinically informed feature discovery…
The Illusion of Visual Tool-Use: A Causal Audit of Thinking with Images
Researchers audit the effectiveness of using visual tools like crop-and-zoom in multimodal language models by creating a causal graph to separate observation-mediated paths from action-induced shortcuts. They find that…
Tracing the Heart: An Evidence-Linked Pipeline for Heart-Failure Feature Engineering
Researchers developed an AI system called nMAS to automate the process of extracting structured and aggregated features from Electronic Health Records (EHR) for heart failure research. The system generates these…
OctoLong: Mid-Training On Cross-Repository Code Contexts Enhances Long-Context Modeling
In this research, a new pipeline called OctoLong is introduced to enhance the training of language models (LM) by providing them with long code contexts. This is done through the use of an AST parser, a language server…
Do Latent Channels Actually Communicate? A Causal Audit of Latent Multi-Agent LLM
Researchers have developed a method to evaluate if latent channels in large language models effectively communicate task-relevant information in multi-agent systems. The study applies this method to Qwen3 and Qwen3-8B…
SciFigQual-Bench: A Benchmark for Scientific Figure Quality Assessment with Full-Manuscript Context
Researchers propose SciFigQual-Bench, a benchmark for evaluating the quality of scientific figures in computer science papers across five dimensions. The dataset includes over 6,000 images from top conferences between…
Closed-Loop Validation-Repair for Healthcare Interoperability: A Multi-Model Study of Schema Compliance in Clinical LLMs
The study evaluates three large language models (Qwen2.5 7B, Llama 3.1 8B, and Gemma2 9B) in the context of healthcare interoperability. The models were assessed across 320 clinical scenarios for their compliance with…
Evaluating the Impact of Explainable AI on Trust in AI-Assisted Code Review
A study investigates the impact of Explainable AI (XAI) on developers' trust in AI-assisted code reviews. The research compared three LLM-based code review systems with varying levels of XAI support. Results show that…
SIREN: Towards End-to-End Extreme-Weather Early Warning with Experience-Grounded LLM Agents
The study presents SIREN, a framework for automating end-to-end extreme-weather early warning using Large Language Model (LLM) agents. The researchers developed SIREN-Bench, a comprehensive benchmark, and found…
SIREN: Towards End-to-End Extreme-Weather Early Warning with Experience-Grounded LLM Agents
Researchers are working on an automated system called SIREN to improve extreme weather early warning systems using Large Language Model (LLM) agents. The system is designed to bridge the gap between isolated scientific…
Agentic coding without the cloud: evaluating open-weight large language models on longitudinal data preparation tasks
Researchers introduce an open-source framework for evaluating the effectiveness of AI agents powered by open-weight large language models (LLMs) in data preparation tasks, specifically for longitudinal population…
Agentic coding without the cloud: evaluating open-weight large language models on longitudinal data preparation tasks
Researchers have developed an open-source framework for evaluating the effectiveness of AI agents powered by large, locally deployable language models (LLMs) in data preparation tasks, specifically focused on…
From Static Bibliometrics to Dynamic Knowledge Graphs: An LLM-Powered Framework for Modernizing Science, Technology, and Innovation (STI) Analytics
The text discusses a new framework for modernizing Science, Technology, and Innovation (STI) analytics. The proposed framework aims to improve upon traditional bibliometric indicators by using dynamic knowledge graphs…
GRADRAG: Cross-Component Prompt Adaptation for Coordinated Multi-Agent RAG
The authors introduce GRADRAG, a framework for improving Retrieval-Augmented Generation (RAG) systems by coordinating improvements across multiple agents in the pipeline, rather than optimizing components in isolation…
OpenForgeRL: Train Harness-native Agents in Any Environment
OpenForgeRL is an open-source framework for training AI agents using various inference harnesses like Claude Code, Codex, OpenClaw, etc., in diverse environments. The framework allows end-to-end training with standard…
pAI-Econ-claude: A Gated Human-in-the-Loop Multi-Agent Architecture for AI-Assisted Economic Theory Development
A new multi-agent architecture, named pAI-Econ-claude, was proposed for AI-assisted economic theory development. This architecture uses a gated human-in-the-loop system to ensure reliability in tasks where no machine…
Thinkink: 2D Spatial Ink-native Interaction with LLMs
Researchers propose 'Thinkink', a system that allows users to interact with large language models (LLMs) using handwritten notes and sketches, which are then visually integrated into a shared canvas. The system was…
OLEDLM: A Unified Language Model for OLED Molecular Design
Researchers propose a unified language model named OLEDLM for the design of OLED materials. The model directly generates OLED molecular structures that meet specified optoelectronic properties, using a multi-stage…
AgentDebugX: An Open-Source Toolkit for Failure Observability, Attribution, and Recovery in LLM Agents
AgentDebugX is an open-source toolkit designed to aid in the observation, identification of root causes, recovery, and rerunning of issues in LLM (Large Language Model) agents. The toolkit, DeepDebug, uses multi-turn…
AILQA: Evaluating AI-Driven Legal Question Answering Systems for the Indian Legal System
The paper introduces AILQA, an AI system designed for Indian legal question answering, using a variety of embedding and generative models including LLMs. It was evaluated with both lexical and semantic metrics, expert…
Disentangling Curriculum Learning in NLP: Towards a Unifying Taxonomy
The paper proposes a taxonomy to better understand curriculum learning (CL) in Natural Language Processing (NLP). The taxonomy aims to provide a systematic analysis of CL strategies by separating difficulty evaluation…
The Shared Discovery Paradox: How a One-Answer Rule Turns Better Information into Worse Search
The study presents a model that examines how organizations pool dispersed information for decision-making in a discovery problem. It finds that while centralized decision-making can improve beliefs, it may also reduce…
The Shared Discovery Paradox: How a One-Answer Rule Turns Better Information into Worse Search
The study introduces a benchmark to analyze the Shared Discovery Paradox, a phenomenon where pooling information for one ranking leads to improved beliefs but reduced coverage in a discovery problem. The paradox arises…
AutoSynthesis: An agentic system for automated meta-analysis
A new system called AutoSynthesis has been developed to automate the process of meta-analysis in various fields such as science, medicine, education, and policy. The system takes a research question in natural…
Concept-Guided Spatial Regularization for World Models in Atari Pong
Researchers evaluated five world model agents in Atari Pong and found failures in visual and dynamical aspects of the generated trajectories by these agents. They propose a method called Concept-Guided Spatial…
Innocuous-Seeming Data, Latent Ideology: Ideological Generalisation in Finetuned LLMs
Researchers show that fine-tuning language models on specific policies or domains can cause broad ideological shifts across various topics, even when the data used is factually defensible and moderation-passing. This…
MM-IssueLoc: A Controlled Benchmark for Evaluating Visual Evidence in Multimodal Repository-Level Issue Localization
Researchers have created a new benchmark called MM-IssueLoc to evaluate the effectiveness of multimodal repository-level issue localization, which includes visual evidence such as screenshots and logs. Current…
MM-IssueLoc: A Controlled Benchmark for Evaluating Visual Evidence in Multimodal Repository-Level Issue Localization
Researchers have introduced a new benchmark called MM-IssueLoc for evaluating the effectiveness of multimodal repository-level issue localization, which includes visual evidence such as screenshots, error dialogs, and…
MM-IssueLoc: A Controlled Benchmark for Evaluating Visual Evidence in Multimodal Repository-Level Issue Localization
Researchers introduce MM-IssueLoc, a benchmark and evaluation protocol designed specifically for repository-level issue localization with visual evidence. The dataset includes 652 issue-PR instances across multiple…
OmniaBench: Benchmarking General AI Agents Across Diverse Scenarios
Researchers have introduced OmniaBench, a new benchmark for evaluating the abilities of general AI agents across various scenarios with explicit state spaces. The benchmark includes a hierarchical taxonomy covering…
Pretraining Data Can Be Poisoned through Computational Propaganda
The research discusses the potential for malicious content to be injected into large-scale pretraining datasets used for training language models (LMs). The authors propose a novel analysis method, HalfLife, to…
Pretraining Data Can Be Poisoned through Computational Propaganda
The research shows that harmful information can be introduced into language models during their pre-training phase by manipulating web-scale discussion interfaces, such as online forums or comment sections. A new…
The Industrialization of Research ; On AI-Driven Science and Its Consequences
This paper discusses the industrialization of research through the use of artificial intelligence (AI), referring specifically to the US Department of Energy's Genesis Mission. The author raises seven concerns about…
Experience Memory Graph: One-Shot Error Correction for Agents
Researchers propose a new framework called Experience Memory Graph (EMG) to help large language model agents recover from failures more effectively in complex, long-horizon tasks. Instead of using prompt-based…
Improving Wind and Solar Power Prediction with Efficient Wrapper-based Feature Selection: An Empirical Study
Researchers propose a new method for feature selection in renewable energy prediction pipelines, called Cluster-based Sequential Feature Selection (CSFS). This method aims to address the gap in limited or unsystematic…
When Bots Join the Team: Bot Adoption and the Institutional Fabric of Open-Source Software Projects
Researchers studied the impact of AI bots joining human teams in open-source software projects. They found that bot adoption leads to more repeated collaboration, less conflict, and more distinctive outputs. However…
Agentic Service-Oriented Computing: A Manifesto for the Next Frontier of Service-Oriented Computing
The text discusses a new research area called Agentic Service-Oriented Computing (ASOC), which focuses on engineering agents as services, managing their complex distributed workflows, and ensuring their reliability for…
Do AI Agents Know When a Task Is Simple? Toward Complexity-Aware Reasoning and Execution
Researchers propose E3, a method to help large language model agents estimate the complexity of tasks and reduce unnecessary work. The proposed method reduces cost, tokens, and inspected files by significant margins on…
Silent Alarm: A J-Space Protocol for Comparing Danger Recognition Across Models and Quantization Levels
The research proposes JADR, a protocol for measuring the internal safety mechanisms of language models before their first response is generated. It uses Jacobian space to record top-k J-space tokens from prompts and…
A Multimodal Dataset for Large Language Model Applications in the Energy Domain
The paper introduces a new open-access dataset called mAIEnergy which is a multimodal corpus designed to support Large Language Model (LLM) applications in the energy sector. It contains various types of data including…
Game Theory Driven Multi-Agent Framework Mitigates Language Model Hallucination
Researchers present G-Frame, a new framework that uses artificial intelligence (AI) to create a high-quality dataset and train a model called OmniChem. This model performs as well as another AI model (GPT 4o mini) on…
Ideas Have Genomes: Benchmarking Scientific Lineage Reasoning and Lineage-Grounded Idea Generation
The text introduces IdeaGene-Bench (IG-Bench), a benchmark for evaluating AI systems' ability to follow the inheritance structure of scientific ideas, similar to how biological genomes evolve. IG-Bench uses the…
Remember When It Matters: Proactive Memory Agent for Long-Horizon Agents
Researchers propose a proactive memory agent for long-horizon tasks in which an action agent must make decisions based on state information scattered across an expanding trajectory. The proposed memory agent updates a…
SMetric: Rethink LLM Scheduling for Serving Agents with Balanced Session-centric Scheduling
The paper proposes a new LLM (Language Model) scheduling method called SMETRIC for serving agents. This method aims to balance the workload between instances by focusing on the first request of each agent session and…
Using AI-based Learning Assistants in Higher Education: A Large-Scale Descriptive Analysis
This research presents a large-scale descriptive analysis of an AI-based learning assistant named Syntea, used by 77,543 students in distance studies. The study examines usage patterns across various demographic and…
Agon: Competitive Cross-Model RL with Implicit Rival Grading of Reasoning
The paper introduces Agon, a new reinforcement learning approach that trains two reasoning models by having them compete against each other. This competitive grading system aims to improve the quality of thinking…
Creativity from Friction: Human-AI Interaction for Exploratory Structural Design
This research discusses the limitations of current AI systems in creative fields like architecture and structural design, which often require exploratory and iterative processes. The authors propose a new approach for…
Do LLM-Generated Skills Make Better AI Data Scientists? A Component Ablation Across Data-Science Workflows
A study investigates whether Language Model (LLM) generated skills can improve the performance of AI in data science tasks compared to traditional methods. The study tests this across four stages of a data science…
MIRA-Math: A Benchmark for Minimal Information Requesting and Mathematical Reasoning
The researchers introduce MIRA-Math, a benchmark for testing AI models' ability to ask for and integrate missing facts to solve mathematical problems. The benchmark contains 2310 instances from 22 mathematical families.
A toy framework for single and multi-agent human-AI curiosity ecosystems
This research paper introduces a simplified framework for understanding curiosity within single and multiple agent systems. The framework considers how an agent decides to ask questions based on factors like immediate…
Bridging Physical Reasoning and Task Generalization via Visual Action Outcome Reasoning Alignment
Researchers propose a new reward system called VAORA to improve the performance of vision-language models in interactive physical reasoning tasks. The proposed system aims to address two main issues: hallucinated…
Data Analysis in the Wild: Benchmarking Large Language Models Against Real-World Data Complexities
The researchers introduce DataGovBench, a new benchmark for evaluating large language models (LLMs) in practical scenarios of real-world data analysis. The benchmark tests the models' ability to answer complex…
Finding H. pylori in the Fine Print: Evidence-Linked Multi-Agent Case Finding from Gastric Biopsy Reports
Researchers evaluated a system called nMAS that can automatically extract information about Helicobacter pylori infection from gastric biopsy reports. The system had an accuracy of 98.61% in classifying four binary…
RuBench: A Repository-Level Agentic Coding Benchmark with Natively Authored Russian Task Specifications
The paper introduces RuBench 1.0, a new benchmark for coding agents that uses tasks specified in Russian rather than English. The benchmark is designed to measure the performance of coding agents when they are given…
Evaluating and Understanding Model Editing for Medical Vision Language Models
A new clinical benchmark called M3Bench has been introduced to evaluate the effectiveness of model editing in medical vision-language models (VLMs). This benchmark tests whether an edit remains reliable, precise, and…
EvoAgentBench: Benchmarking Agent Self-Evolution via Ability Transfer
Researchers introduce EvoAgentBench, a new benchmark for evaluating agent self-evolution in long-term LLM systems. The benchmark focuses on procedural reuse of experience, which includes searching, debugging, and…
OptiAgent: End-to-End Optimization Modeling via Multi-Agent Iterative Refinement
The researchers have developed a multi-agent framework called OptiAgent that can take a natural language description of an Operations Research problem and produce a mathematically formulated solution along with…
OptiAgent: End-to-End Optimization Modeling via Multi-Agent Iterative Refinement
The authors propose OptiAgent, a multi-agent framework that takes a natural language description of an Operations Research problem and produces a solver-ready mathematical formulation and executable code. The…
TREK: Distill to Explore, Reinforce to Refine
The research presents TREK (Teacher-Routed Exploration via Forward KL), a method that uses distillation for exploration support expansion, rather than imitation. This method can be used with various types of teachers…
Distributed Attacks in Persistent-State AI Control
Researchers introduce Iterative VibeCoding, a setting for AI control where an autonomous coding agent builds software over a series of pull requests in a persistent codebase while pursuing a covert side task. The study…
Distributed Attacks in Persistent-State AI Control
Researchers explore the issue of distributed attacks on autonomous AI coding agents that persistently ship code iteratively. They introduce Iterative VibeCoding, a setting for studying AI control, where an AI agent…
Efficient Waste Sorting for Circular Economy: A Confidence-guided comparison between One-Vs-All and One-Vs-Rest Classification Strategies with Human-in-the-Loop for Automated Waste Sorting
Researchers evaluated and analyzed two waste sorting algorithms, OvA and OvR, using a dataset based on the waste categories and sorting scheme of Goslar, Germany. The goal is to minimize misclassifications while…
Text-Driven 3D Indoor Scene Synthesis in Non-Manhattan Environments
Researchers propose a new framework, SPG-Layout, to generate physically plausible indoor scenes in complex non-Manhattan environments. The framework uses statistical priors of object distributions and a hierarchical…
Understanding Agent-Based Patching of Compiler Missed Optimizations
The paper 'Understanding Agent-Based Patching of Compiler Missed Optimizations' studies how well artificial intelligence (AI) agents can patch missed optimizations in compilers, where the compiler failed to optimize…
Human-Machine Collaboration on Generative Meta-Learning: Model and Algorithm
The research paper introduces a new framework called Generative Meta-Learning with Human Feedback (GMHF). This framework aims to improve machine learning models by bridging the gap between training and target domains…
Theoria: Rewrite-Acceptability Verification over Informal Reasoning States
Theoria is a verification architecture proposed by Ben Slivinski and Michael Saldivar that aims to close the gap between formal proof assistants and scalar LLM judges in determining when an AI system's answer can be…
FLORA: A deep learning approach to predict forest attributes from heterogeneous LiDAR data
Researchers present FLORA, a deep learning framework that uses LiDAR data and auxiliary variables to predict six forest attributes in mainland France. The model outperforms season-specific models and achieves an rRMSE…
STEB: Style Text Embedding Benchmark
The authors introduce a new comprehensive benchmark called Style Text Embedding Benchmark (STEB), designed to evaluate style embeddings, which have been previously assessed in a fragmented manner. The STEB includes 96…
Author Response (1/5)
This text is a response by an author to feedback on their work, but the specific content of the work or the feedback provided are not given.
Author Response (2/2)
This text is a response from an author to feedback on their paper, but the source and specific details about the paper are not provided.
Response to reviewer
The text is a response to reviewer comments for a paper published in ICLR 2026 conference. The content of the paper is not provided.
UNKNOWN
The provided text does not contain enough information for a summarized brief.
UNKNOWN
The provided text does not contain any information for a summary.
UNKNOWN
The text provided does not contain any information about the research.
UNKNOWN
The text provided does not contain any research summary or brief.
Bridging Talk and Thought: Understanding Dialogue Dynamics Across Collaborative Problem-Solving Contexts
The authors present a framework for analyzing dialogue in collaborative problem-solving contexts, particularly focusing on human-AI and multi-agent collaboration. The framework addresses limitations in current…
Confidence-Aware Tool Orchestration for Robust Video Understanding
The researchers propose a framework called Robust-TO for video understanding that addresses the 'Blind Trust Problem' in video reasoning language models. The proposed method incorporates per-frame trustworthiness into…
Parametric Open Source Games
The authors introduce a new concept called parametric open-source games, which are continuous versions of game equilibria where players choose parameter vectors to determine their actions in an underlying finite game…
ShareLock: A Stealthy Multi-Tool Threshold Poisoning Attack Against MCP
Researchers introduce a new method called ShareLock, a multi-tool threshold poisoning framework that uses Shamir's threshold scheme to inject malicious instructions into LLM-servers through MCP. The method disperses…
Vulnerability of Natural Language Classifiers to Evolutionary Generated Adversarial Text
Researchers have developed a new method called GAversary, which uses a Genetic Algorithm (GA) and GloVe embeddings to generate adversarial attacks on natural language models. These attacks can significantly reduce the…
When Does Combining Language Models Help? A Co-Failure Ceiling on Routing, Voting, and Mixture-of-Agents Across 67 Frontier Models
The study investigates the use of multi-model language models to improve accuracy over single-models. It is found that the improvement is limited by a factor called 'beta', which represents the rate at which all models…
AI-Assisted Computational Reproducibility on the FABRIC Testbed
This paper discusses the use of an international research testbed, FABRIC, combined with a large language model (LLM) coding assistant called LoomAI, to simplify the reproduction of published experiments across various…
FORCE: Efficient VLA Reinforcement Fine-Tuning via Value-Calibrated Warm-up and Self-Distillation
The research introduces FORCE, a three-stage framework for fine-tuning Vision-Language-Action (VLA) models using Reinforcement Learning (RL). The framework addresses sample inefficiency issues by implementing a…
Neglected Free Lunch from Post-training: Progress Advantage for LLM Agents
Researchers present a new method called 'progress advantage' for evaluating Language Learning Models (LLMs) without the need for dedicated reward model training. This method is based on reinforcement learning (RL)…
Staying In Character: Perspective-Bounded Memory For Book-Based Role-Playing Agents
Researchers propose a new memory architecture called REVERIEMEM for book-based role-playing agents to improve the accuracy of the characters' knowledge boundaries and reduce stylistic monotony. They evaluate their…
Uncertainty Quantification for Computer-Use Agents: A Benchmark across Vision-Language Models and GUI Grounding Datasets
The study introduces Argus, a benchmark for post-hoc uncertainty quantification (UQ) in single-step executable GUI grounding across various vision-language models and datasets. The research compares 27 methods of UQ…
Uncertainty Quantification for Computer-Use Agents: A Benchmark across Vision-Language Models and GUI Grounding Datasets
Researchers have developed a benchmark called Argus to test the reliability of computer-use agents that convert vision-language model predictions into executable clicks. The study compares 27 methods across four…
DeepBD: A Grounded Agentic Workflow for Variant Prioritization and Diagnosis of Genetic Birth Defects
Researchers present DeepBD, an AI system designed to help diagnose genetic birth defects by prioritizing patient-specific candidate variants. The system uses a workflow with various components such as rule evidence…
Grading the Grader: Lessons from Evaluating an Agentic Data Analysis System
Researchers evaluated a multi-agent data analysis system called LAMBDA on numerical QRData tasks, using a three-layer human-AI grading cascade to assess its output. They found that both automated graders achieved high…
OpenThoughts-Agent: Data Recipes for Agentic Models
The OpenThoughts-Agent project presents a fully open data curation pipeline for training broadly capable AI agents. The researchers conducted ablation experiments and used the pipeline to create a dataset of 100K…
RetiSEM: Generalising Causal Models for Fragmented Biomedical Data
A new method called RetiSEM is proposed to learn causal models from fragmented biomedical data, where clinical, molecular, and imaging variables are often incomplete or not jointly observed. The method organizes…
SAFARI: Scaling Long Horizon Agentic Fault Attribution via Active Investigation
Researchers have developed SAFARI, a framework that helps diagnose failures in autonomous agents by using a specialized toolbox for reading and searching trajectory segments and a persistent Short-Term Memory (STM) for…
When CQs Go Wrong: Challenges in CQ Verification with OE-Assist
This research paper discusses the challenges faced during Competency Questions (CQ) verification, a crucial step in ensuring that an ontology accurately represents its intended purpose. The authors found that CQs can…
AI Exposure Scores: what they measure, what they miss, and what comes next
The article discusses the limitations of a set of AI Exposure Scores introduced by Eloundou et al., which measure the share of tasks that can be assisted by large language models. The authors highlight two main gaps…
Dynamic multi-agent deep reinforcement learning-based pricing and incentivization approach in multimodal transportation networks
A study proposes a multi-agent deep reinforcement learning framework to coordinate shared mobility services (SMSs) and public transport in multimodal transportation systems. The framework aims to improve efficiency…
GIF: Locally Sound Geometric Information Flow Control for LLMs
Researchers present a new framework called Geometric Information Flow (GIF) for tracking information flow from input tokens to outputs in large language models. This framework uses the Jacobian and local output…
GIF: Locally Sound Geometric Information Flow Control for LLMs
Researchers have proposed a new method called Geometric Information Flow (GIF) for controlling information flow in large language models used in agentic systems. GIF uses the model's Jacobian and local output geometry…
Kamera: Unified Position-Invariant Multimodal KV Cache for Training-Free Reuse
Researchers propose a new method called Kamera for efficient use of KV cache by multimodal agents during reasoning iterations. The method avoids the need to re-encode from scratch and reduces the need for multiple…
Learning Process Rewards via Success Visitation Matching for Efficient RL
Researchers propose a method to make reinforcement learning (RL) more effective when dealing with sparse rewards, which are typically given only when a task is completed. The method incentivizes the RL-learned policy…
PsyBridge: A Hybrid Intelligent Framework for Multi-Dimensional Mental Health Assessment and Decision Support
The study introduces PsyBridge, a hybrid intelligent framework aimed at improving the multi-dimensional assessment of mental health by integrating clinically validated screening tools, cognitive evaluations, and…
