Continual harness evolution causes forgetting in coding agents
Meta-Harness drops 66% on SWE-Bench after evolution on FeatureBench with DeepSeek-V4-Flash-0731.
Stories per day · 30 days
3 RSS
Relatedagent-harness2agentmark1catastrophic-forgetting1claude-code1coding-agents1continual-evolution1counterfactuals1gepa1physical-reasoning1provenance1qwen1simulation1vlm1watermarking1
13 stories, newest first · one card per story
Stories tagged #self-evolution
PhysEvolve trains VLMs on executable worlds for physical reasoning
From Qwen3-VL-8B-Instruct, the paper reports PhysBench Throwing up 8.8 and ContPhy Goal-Pit up 24.4 points.
What Survives Self-Evolution: Provenance Marks in Copied Agent Harnesses
Provenance marks in leaked agent harnesses mostly fail when a self-evolving LLM rewrites the marked code.
Self-Evolving Agent Harness
Paper formalizes persistent self-evolution of agent harnesses with fixed weights and validation bounds.
Skill adaptation improves weak models in harness self-evolution
Across 103 task-skill pairs, skills that raise GPT-5.6-luna by 31.1 points lower Gemma-4-E4B by 8.8.
Capability-Driven Self-Evolution of Agent Memory
PrisMem evolves agent memory by capability and beats baselines on long-history benchmarks.
Do Self-Evolving Skills Generalize to Held-Out Tasks?
Of 21 self-evolving skills that improve in training, only five fully transfer their gains to held-out tasks.
Learning from Runtime Feedback through Failure-Bank Self-Evolution for Vision-Language-Action Models
FailBank converts runtime safety feedback into persistent policy gains for vision-language-action robot models.
False Frontiers: Diagnosing and Mitigating Co-Cheating in Self-Evolving Search Agents
CrossFit cuts co-cheating in self-evolving search agents and beats coupled self-evolution on seven benchmarks.
Program-Verified Self-Evolution for Vision-Language Models
VQS generates program-verified vision questions, labeling 94% correctly and lifting Qwen3-VL by up to 3.84 points.
TimeEvo: Failure-Driven Self-Evolution of a Time Series Agent
TimeEvo builds a time-series agent tool library from diagnosed failures and raises accuracy on ten QA tasks.
Procedural Graphs: Self-Evolving Execution Structures for LLM Agents
Researchers introduce Procedural Graphs, self-evolving (procedure, relation, procedure) structures guiding LLM agent tool use and planning.
StudyBench: Can Self-Evolution Squeeze Textbooks for Olympiad Capability?
Researchers introduce StudyBench, a physics benchmark showing self-evolution gains on textbook problems rarely transfer to olympiad-level questions.