微软研究：清空对话历史可让 AI 智能体找到更好方案

Oct 12, 2026

Image description

Nano Banana Pro prompted by THE DECODER

Long-running AI research agents can plateau despite additional compute, a Microsoft Research study finds. Accumulated conversation context can constrain their search, keeping them focused on similar solutions instead of exploring better approaches.

Microsoft researchers studied what they call "Autoresearch agents," AI systems that repeatedly propose solutions, evaluate them, and refine them based on feedback. Eight independent runs of the same agent on the same task plateaued at significantly different performance levels, with final scores ranging from 68.3 to 100.0. These gaps persisted even with substantially more compute, according to the paper.

Agents get stuck in "idea basins"

The researchers plotted solutions from all runs on a two-dimensional map that places similar solutions close together, then compared them. Solutions from a single run almost always stayed within a small region of the map. The authors call these regions "idea basins."

The agents spend a large share of their remaining compute budget refining solutions within a less productive region instead of finding better approaches discovered by other runs. This pattern persists even when the agents periodically compress their conversation histories.

Clearing chat history lets agents explore new approaches

The proposed method, "Fork-and-Flush," combines two mechanisms. In "forking," the agent splits into K parallel branches, each inheriting the previous workspace and all generated files. In "flushing," each branch loses its entire chat context but keeps the workspace, allowing it to start a fresh conversation and build on prior work.

On the left, a single run (gray, dashed) keeps refining solutions within the same idea basin. Fork-and-Flush starts parallel branches with fresh chat contexts, keeps the best branch, and discards the rest. On the right, a branch with fresh context moves into a different idea basin and achieves a higher score. | Image: Cai, Ziakas et al. (2026)

A branch that keeps the parent process's chat context tends to return to the same idea basin because the previous reasoning steers its next steps. Fresh context restores diversity in the search. The conversation history is lost, but the code, files, and saved results remain in the workspace.

A 66 percent higher normalized average score on the same compute budget

The researchers tested Fork-and-Flush on 13 long-running tasks, including polyomino packing, job-shop scheduling, and cryptanalysis. The tasks also included compressing a CIFAR classifier and optimizing a VLIW kernel. Individual runs lasted from several hours to over 100 hours. The agents used GPT-5.6-terra and GPT-5.5 through the GitHub Copilot CLI.

On the same compute budget, Fork-and-Flush achieved a min-max-normalized average score of 0.78, compared to 0.47 for single runs and 0.54 for best-of-4. The paper reports relative gains of 66.0 and 44.4 percent, respectively. The method achieved the highest mean score on 12 of the 13 tasks.

Fork-and-Flush also produced more consistent results, with normalized standard deviation falling to 0.16 from 0.26 for single runs. Simply clearing the chat context without branching improved results by 22 percent.

Evaluation	Method	Normalized Average Score	Normalized Standard Deviation	Relative improvement over the respective single-run baseline
13 tasks	Single run	0.47	0.26	Baseline
13 tasks	Best-of-4	0.54	0.21	approx. +14.9%
13 tasks	Fork-and-Flush	0.78	0.16	+66.0%
11 tasks	Single run	0.45	—	Baseline
11 tasks	Flushing only	0.55	—	approx. +22%

Stronger models may still need context resets

The researchers also tested whether more capable models avoid idea basins, using MAX-3-SAT problems across three model sizes, Qwen3.5-4B, 9B, and 27B. Variance between runs followed an inverted U-shaped curve, peaking on tasks at the edge of a model's abilities, roughly halfway between random-guessing performance and the reference solver's performance.

Results vary most on tasks at the edge of a model's abilities, with some runs succeeding and others falling well short. Models almost always succeed on very easy tasks and fail on very hard ones. This pattern holds across all three model sizes tested. | Image: Cai, Ziakas et al. (2026)

The authors argue that open research problems sit, by definition, at the limits of models' abilities. They expect the variation Fork-and-Flush exploits to persist as models improve, with the boundary simply shifting toward harder problems. The results suggest that regular context resets could remain useful even for frontier research agents.

AI News Without the Hype – Curated by Humans

Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.

Subscribe now

Full access to every article on THE DECODER
No ads
Join the comments and community discussions
A weekly AI news recap via mail
6x/year: "AI Radar" — deep dives on the AI topics that matter most
Daily AI news, always up to date
Our full ten-year archive
Covered by a team with 10+ years in AI