GEA: Letting Agents Share Evolutionary Gains
xwang_lk · x · 2026-07-15
The author shares research on GEA, which attempts to break away from the biological lineage-based evolutionary approach for self-evolving agents. It emphasizes that agents can directly share experiences and artifacts without being constrained by reproduction, lineage, or genes.
Key results include achieving 71.0% on SWE-bench Verified and 88.3% on Polyglot with zero human intervention. The work has been accepted by COLM 2026. The core argument is that traditional tree-based evolution creates many short-lived branches; while exploration happens, there is insufficient reuse and accumulation. GEA aims to more directly consolidate and reuse exploration outcomes.
More from coding & agent
- HeyGen adds a media-sourcing skill for coding agents with 75k images and 10k tracks — HeyGen · 2026-07-22
- Agent search bottlenecks are now about variance, not raw latency — rohanpaul_ai · 2026-07-22
- LangSmith adds tracing for Pipecat, LiveKit, OpenAI Realtime, and Gemini Live — LangChain · 2026-07-22
- An MCP server signs every AI agent tool call into a verifiable Merkle chain — Funky_Chicken_22 · 2026-07-22
- Annotated transcript of a Claude Code team interview is now available — trq212 · 2026-07-22
- Claude Code skill uses 10 Markdown rules to make outputs ADHD-friendly — alex_verem · 2026-07-22