Papers

BRED:用 LLM 控制器做自主研究的轮间决策

The Between-Round Decision in Autonomous Research: An LLM Controller That Reads the Whole Trajectory

Why we picked it

这论文回答了个很实际的问题:自主研究系统跑完一轮实验后,下一步该继续调优还是换方向。BRED 不用搜索树和打分规则,直接让 LLM 读完整轨迹来做决定,在 ETTh1、SST-2、CIFAR-100 上都超过了已发表最好成绩。做 agent 框架或自动化实验的人都该看看实现思路。

arXiv 论文 2611.22661 提出 BRED(Between-Round Evidence-based Decision),一个用 LLM 做轮间决策的控制器:每轮实验结束后,由它决定是继续优化当前最优解(Champion)还是开辟新方向、基于哪个方案、设什么目标,不依赖规则、评分、搜索树等外部结构。

Body

在三个常用 research-agent 基准上,BRED 超越已发表结果:ETTh1 的 MAE 降至 0.4172(提升 4.80%),SST-2 准确率 95.07%(+3.44),CIFAR-100 准确率 84.63%(+3.43)。消融实验显示证据阅读是关键:不读历史时 ETTh1 MAE 为 0.4299、CIFAR-100 为 83.71%,完整控制器分别再低 2.95% 和高 0.92 个点。

在新增的语音抑郁检测任务(无泄漏协议)上,BRED 在 held-out 种子上达到 macro-F1 0.6164,比基线高 5.5 分,比最强已发表系统高 4.0 分。代码已在 GitHub(learninginvision/BRED)开源。

Original · arXiv cs.AI

The Between-Round Decision in Autonomous Research: An LLM Controller That Reads the Whole Trajectory

Executing an idea often changes how promising it looks, so the next step in autonomous research should rest on results, not on proposals. After each round of experiments, the system must decide whether to refine the best solution or start a new direction, which solution to build from, and what goal to set. We call this the between-round decision. Existing systems wrap it in machinery such as rules, scores, search trees, and schedules; where an LLM makes it, what the LLM needs to read is left open. We ask whether an LLM can make this decision from accumulated evidence, without a search structure around it, and how much of that evidence it needs to read. We present BRED (Between-Round Evidence-based Decision), an LLM-based controller that maintains the best verified solution so far (the Champion) and a log of past rounds, and makes one decision per round. On three widely used research-agent benchmark tasks, BRED improves on the task baselines by 4.80% in MAE on ETTh1 (to 0.4172) and by 3.44 and 3.43 points of accuracy on SST-2 (to 95.07%) and CIFAR-100 (to 84.63%), ahead of the published results. When the controller does not read the accumulated evidence, the same loop gains less: with no history it reaches 0.4299 on ETTh1 and 83.71% on CIFAR-100, whereas the full controller's MAE is 2.95% lower and its accuracy 0.92 points higher. On a new task, speech-based depression detection under a leakage-free protocol, it reaches a macro-F1 of 0.6164 on seeds held out from selection, 5.5 points above its baseline and 4.0 above the strongest published system. Code is available at https://github.com/learninginvision/BRED.

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Papers" desk. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include BRED, research-agent, AI for research, agents.

Source attribution: curated from arXiv cs.AI (@Caoxin Wang, Shengkai Gan, Xiang Hu, Lili Pan).