ScholarPulse 日报 2026-10-07

2026-10-07 学术简报：2 篇。本文提出对抗性启发式学习（AHL）范式与AAArena基准，证明固定权重的AI智能体仅靠修正可执行策略，即可在部分对抗游戏中登顶人类天梯。

今日速览

序号	标题	来源	日期	主题	推荐等级
1	Can AI Agents Learn Their Way to the Top? Evaluating Heuristic Learning in a Long-Running Game Agent Competition	arXiv	2026-10-08	AI-Agent	高
2	DataSense-Bench: The First Step Toward an AI Scientist	arXiv	2026-10-08	AI-Agent	高

重点论文与技术动态

1. Can AI Agents Learn Their Way to the Top? Evaluating Heuristic Learning in a Long-Running Game Agent Competition

来源：arXiv
日期：2026-10-08
作者/机构：Kaisen Yang, Qingle Liu, Kejin Wang, Yicheng Zhao, Jieming Li, Shenghan Zheng
主题标签：AI-Agent, arXiv
推荐等级：高
分类：cs.AI, cs.CL

一句话结论

本文提出对抗性启发式学习（AHL）范式与AAArena基准，证明固定权重的AI智能体仅靠修正可执行策略，即可在部分对抗游戏中登顶人类天梯。

核心内容

提出AHL范式，以AI智能体为学习引擎，在保持模型权重固定的前提下，将游戏经验转化为策略及软件代码的修正。
评估显示，Opus5.5配合Claude Code在12款游戏中斩获6枚金牌，但在其余6款及规则更复杂的游戏中表现较弱，未能登顶。
实验证实，对手选择、密集反馈以及自身与他人的比赛回放（同策略与异策略）均能有效支持策略改进。

方法与数据

构建含12款真实对抗游戏和1920个人类程序的AAArena基准，智能体在固定预算内解读规则、选对手、看回放并修改代码；具体评估的模型配置总数摘要未明确。

价值判断

值得关注：无需更新模型权重，仅靠代码修订就能在半数测试游戏中击败人类顶尖程序。
可复用点：结合对手选择、密集反馈及多源比赛回放来迭代优化可执行策略的机制。
局限/待核查：在复杂规则下的游戏理解、策略实现及长视距策略开发上仍面临持续挑战。

摘要

Adversarial games have driven advances from heuristic search to reinforcement learning, yet learning and adapting strategies from limited samples remain challenging. AI agents offer an alternative by turning game experience into revisions of executable policies. Building on heuristic learning (HL), we formalize Adversarial Heuristic Learning (AHL), a paradigm that uses AI agents as learning engines to refine game policies and supporting software while keeping model weights fixed. We introduce AAArena, a benchmark comprising 12 authentic adversarial games and 1,920 archived human programs, with an evaluation protocol modeled on real-world game competitions. Agents interpret rules, choose opponents, analyze replays, and revise game agents to achieve their highest ranking within fixed match and evaluation budgets. We evaluate \val{completedmodels} model and harness configurations: Opus5.5 with Claude Code earns 6 gold medals, while no evaluated configuration tops the remaining 6 human ladders. Performance is generally weaker in games with more complex rule specifications. Further experiments show that opponent selection and dense feedback support policy improvement, and that agents learn from both on-policy replays of their own matches and off-policy replays of other players' matches. These results highlight HL's potential in adversarial games and identify persistent challenges in game understanding, strategy implementation, and long-horizon policy development.

2. DataSense-Bench: The First Step Toward an AI Scientist

来源：arXiv
日期：2026-10-08
作者/机构：Yudi Zhang, Mingyu Cao, Lu Yin, Mykola Pechenizkiy, Shiwei Liu
主题标签：AI-Agent, arXiv
推荐等级：高
分类：cs.LG

一句话结论

本文提出DataSense-Bench，旨在评估AI智能体挑选微调数据并预测性能的“数据感知”能力，发现当前模型相比随机选择提升有限且跨任务表现不稳定。

核心内容

智能体仅可通过查数据、跑分析代码和模型前向传播（禁止训练模型或访问评估任务），挑选并排序用于微调小模型的候选数据子集。
采用双维度评估：榜首子集的微调后性能（反映识别高价值数据能力）与排序准确率（反映预测相对性能能力）。
实验表明智能体相比随机选择收益有限、排序不可靠且跨任务不一致（如Astra在工具任务3次全中，终端任务仅中1次），对相似数据信号的价值解读也存在差异。

方法与数据

涵盖终端问题解决和工具使用两类任务，分别从OpenThoughts-Agent和EnvScaler中选取轨迹数据，并在TBLite和BFCL基准上进行微调后评估。

价值判断

值得关注：直击AI自我改进（RSI）的核心前提，检验前沿模型是否真正具备数据价值感知力。
可复用点：限制直接训练权限、结合最佳子集性能与排序准确率的标准化评估协议。
局限/待核查：摘要未明确被微调的基座小语言模型具体型号。

摘要

As claims about recursive self-improvement (RSI) and artificial general intelligence (AGI) proliferate, we ask a simple question: do frontier AI models have a sense of data, i.e., can they reliably select the right data for training? We introduce DataSense-Bench to study this capability through the fundamental problem of data selection and performance forecasting in machine learning. We ask AI agents to select and rank candidate training subsets that can be used to fine-tune a small LLM model. Agents are allowed to inspect the data, write and execute analysis code, and run model forward passes, but can not train the model or access the actual evaluation tasks. We then fine-tune the base model on each selected subset and evaluate its post-training performance under a standardized protocol. We instantiate the benchmark in terminal problem solving and tool use, selecting trajectories from OpenThoughts-Agent and EnvScaler and evaluating on TBLite and BFCL, respectively. We then evaluate the agents along two complementary dimensions: the post-training performance of the top-ranked subset, reflecting the ability to identify high-value training data, and ranking accuracy, reflecting the ability to predict the relative performance of the selected subsets. In our experiments, selection gains over random selection are limited; agents do not reliably rank their selected groups, and ranking ability does not hold consistently across tasks: Astra identifies the best group in all three tool-use runs but in only one of three terminal runs. Analysis of execution traces on both tasks shows that agents often use similar data signals while interpreting their training value differently.