ByteDance Paper: Keeping Messy History Outperforms Neat Summaries
rohanpaul_ai · x · 2026-08-31
A new ByteDance paper proposes Chain-of-Experience, suggesting that for test-time improvement, keeping the messy history of attempts works better than summarizing it into memory.
Key Findings:
- Across math, coding, and knowledge benchmarks, self-feedback achieved 71.0% accuracy, beating iterative solving without feedback (66.8%).
- With correctness or executor feedback, accuracy reached 79.3%.
- Overall improvement of 5.6% with 19% lower API costs.
- No weight changes are involved, making this contextual adaptation.
Takeaway: For agents, preserve the trajectory and add reliable feedback; compress only when safe.
More from coding & agent
- Fintech HR built a sourcing agent on 60,000 past applicants — 30 qualified candidates in a month — G0NZ0BOY · 2026-10-16
- Orchestrating a 100-image portrait of humanity for aliens with ChatGPT, Argon and p5.brush — dx8xb · 2026-10-16
- 3x DGX Spark cluster runs GLM-5.3-Flash locally behind a proxy, 1.75M tokens/day at 119W — Teknium · 2026-10-16
- Builder's one rule for trusted desktop agents: read freely, but sending, deleting, buying always ask first — No_Fan5310 · 2026-10-16
- GPT-built unifi-simulator makes MikroTik switches appear inside UniFi Network — Pytlicek · 2026-10-16
- Stanford drops 45-min SpaceX lecture on how Grok Bot's always-on agent harness works — Hesamation · 2026-10-16