The Morning
From the arXiv
Agent Error Dataset: Scaling 50,000 Error--Diagnosis Pairs for Failure Analysis and Error-Aware Post-Training
his paper introduces the Agent Error Dataset (AED), a large collection of 50,000 error-diagnosis pairs for text-based LLM agents. The core method involves a five-stage pipeline (AET) to automatically identify agent failures, diagnose their causes, and propose corrections using existing execution traces. The main contribution is providing a scalable resource for analyzing agent failures and developing error-aware post-training techniques, enabling agents to learn from their mistakes without re-running entire rollouts.
PhantomEnvironments: Training LLM Agents in Fictional Worlds
This paper introduces PhantomEnvironments, a novel method for training LLM agents using purely rule-based, fictional environments. This approach bypasses the need for costly human data or LLM-generated environments prone to hallucinations. The core contributio…
PivotOPD: Learning to Recover from Pivotal Mistakes in Multi-Turn Agents
PivotOPD addresses the compounding errors in multi-turn language agents by identifying and correcting "pivotal mistakes" – actions that significantly derail task progress. The core method involves an on-policy distillation framework that trains agents to both …
Debias It Yourself: Teaching LLMs Cognitive Bias Mitigation Interventions
This paper introduces "Debias It Yourself" (DIY), a framework that translates human cognitive bias mitigation techniques into methods for large language models. The core method involves applying these interventions through in-context examples, instruction tuni…
JuryFlow: Disagreement-Guided Human-in-the-Loop Multi-Agent Evaluation
JuryFlow addresses the unreliability of LLM judges by treating inter-judge disagreement as a valuable signal. Its core method decomposes responses into atomic claims, uses heterogeneous judges to provide per-claim verdicts, and builds a disagreement graph. The…
Persistent Context Graphs for Efficient Memory Compaction in LLM Agents
This paper introduces ReCAP, a novel method for efficient memory compaction in LLM agents. ReCAP constructs a persistent context graph storing attention-derived importance scores a…
Cogentic: Multi-Agent Orchestration for Automated Proof Discovery
Cogentic is a multi-agent system designed for automated proof discovery on complex, open research problems. Its core method involves an iterative prove-verify loop where an orchest…
ComputerSD: Online Self-Distillation from Real-Time Feedback for Computer-Use Agents
ComputerSD addresses the challenge of sparse rewards in training computer-use agents by introducing online self-distillation. It leverages real-time feedback from GUI transitions t…
cua-speedrun: Standardized Benchmarking of the Speed of Computer-Use Agents
This paper introduces `cua-speedrun`, a standardized benchmarking framework for evaluating the speed and efficiency of computer-use agents (CUAs). It addresses the reproducibility …
EviRover: Reinforcing Agentic Perception Beyond a Glance
This paper introduces EviRover, a novel agentic perception system that overcomes the limitations of single-glance visual analysis. Its core method involves training an agent to int…
The Town Square
Google announced Gemini 4 Argon, a new family of AI models that are significantly more capable and efficient, with a focus on multimodal reasoning and expanded context windows.
Workshops
VoiceStudio is an open-source, local alternative to ElevenLabs, offering voice cloning, design, video dubbing, dictation, transcription, and audiobook creation across 646 languages.
Openrig is a multi-agent harness that integrates Claude Code and Codex, allowing them to function collaboratively as a single system.
