The Morning
From the arXiv
ACE: Self-Evolving LLM Coding Framework via Adversarial Unit Test Generation and Preference Optimization
CE is a self-evolving LLM coding framework that addresses the limitations of existing methods by using a solver-adversary architecture. It leverages adversarial unit test generation, where a single LLM generates both candidate code and test inputs designed to trigger execution failures. This execution-centric supervision allows the model to actively discover and correct its own errors, leading to more robust code generation without relying on large annotated datasets.
ATLAS: A Multi-LLM Training Framework for EvoDPO with Adaptive Reference Evolution
ATLAS is a multi-LLM training framework that enables an active agent to self-evolve its policy through collaborative training by specialized meta-agents. Its core contribution is the EvoDPO algorithm, which overcomes limitations of fixed reference models by ad…
Blind Spots in the Guard: How Domain-Camouflaged Injection Attacks Evade Detection in Multi-Agent LLM Systems
This paper introduces "domain-camouflaged injection attacks," where malicious prompts mimic the target document's vocabulary and authority to evade detection. The core contribution is identifying and quantifying the "Camouflage Detection Gap" (CDG), demonstrat…
CentaurEval: Benchmarking Human-in-the-Loop Value in Agentic Coding
CentaurEval introduces a novel benchmark for evaluating human-in-the-loop coding agents by creating "Collaboration-Necessary" problems that are too difficult for humans or LLMs alone. Its core contribution is a standardized framework that dynamically generates…
CODE-SHARP: Continuous Open-ended Discovery and Evolution of Skills as Hierarchical Reward Programs
CODE-SHARP is a framework that uses Foundation Models to autonomously discover and evolve a library of Python programs (SHARPs). These SHARPs encode skills as hierarchical reward programs, where each program defines a success condition and relies on previously…
Cumulative Reasoning with Large Language Models
This paper introduces Cumulative Reasoning (CR), a structured framework that enhances Large Language Model (LLM) problem-solving by mimicking human iterative thought. CR uses LLMs …
Enhancing Causal Reasoning in Large Language Models: A Causal Attribution Model for Precision Fine-Tuning
This paper introduces a causal attribution model that uses "do-operators" to create interventional scenarios, allowing for systematic quantification of LLM component contributions …
General Agentic Planning Through Simulative Reasoning with World Models
This paper proposes that general agentic planning requires "simulative reasoning" within a world model, contrasting it with current reactive decision-making. The core method is to …
GROW: Aligning GRPO with State-Action Modeling for Open-World VLM Agents
This paper introduces GROW, a reinforcement learning framework for open-world VLM agents. GROW addresses limitations of existing methods by decomposing full trajectories into state…
Insights Generator: Systematic Corpus-Level Trace Diagnostics for LLM Agents
This paper introduces the Insights Generator (IG), a novel system for systematically diagnosing failures in LLM agents at a corpus level. IG addresses the limitations of manual ins…
The Town Square
YouTube will now automatically label videos that use AI to generate or alter content, aiming to provide viewers with greater transparency.
Workshops
This repository provides an agent harness for performance optimization, featuring skills, instincts, memory, security, and research-first development for various AI coding tools like Claude Code and Codex.
Taste-Skill enhances AI by providing it with "good taste," preventing the generation of boring and generic content.
