ScholarPulse 日报 2026-10-11

2026-10-11 学术简报：2 篇。本文提出ReTeach自蒸馏框架，仅靠模型自生成尝试与结果级验证，通过多轮反思与重试构建自教师，在保持单次推理的同时提升了推理表现。

今日速览

序号	标题	来源	日期	主题	推荐等级
1	ReTeach: Building a Self-Teacher through Multi-Round Reflection and Retry	arXiv	2026-10-08	RAG	高
2	SSCBench: Evaluating the Evidential Validity of Fault-Injection Tests for Tool-Using LLM Agents	arXiv	2026-10-08	AI-Agent	高

重点论文与技术动态

1. ReTeach: Building a Self-Teacher through Multi-Round Reflection and Retry

来源：arXiv
日期：2026-10-08
作者/机构：Yafeng Tang, Hao Li, Hongsheng Yu, Qiang Fu
主题标签：RAG, arXiv
推荐等级：高
分类：cs.AI

一句话结论

本文提出ReTeach自蒸馏框架，仅靠模型自生成尝试与结果级验证，通过多轮反思与重试构建自教师，在保持单次推理的同时提升了推理表现。

核心内容

构建自教师：对失败尝试交替进行显式反思与重试直至成功或耗尽预算，无需参考答案、外部诊断反馈或跨样本记忆。
结果感知加权：区分初始正确、反思修正和未解决的样本，为类别归一化的蒸馏损失分配独立权重。
同策略蒸馏：学生在自身轨迹前缀上匹配教师的词元级预测分布，将多轮修正收益迁移至单次推理。

方法与数据

在涵盖数学推理、科学问答和工具使用的6个基准上（具体数据集名称摘要未明确），平均准确率较GRPO提升1.39个百分点。

价值判断

值得关注：摆脱对参考答案或丰富外部反馈的依赖，仅靠结果验证实现自蒸馏。
可复用点：按样本最终解决状态对蒸馏损失进行分类归一化与差异化加权的策略。
局限/待核查：采用的基础模型及重试预算的具体数值摘要未明确。

摘要

Self-distillation can improve reasoning without a separately trained, more capable teacher, but its effectiveness depends on how the self-teacher gains an advantage over the student. Conditioning the teacher on reference answers or solutions can provide such an advantage, but this information may be unavailable. Reflection offers a way to derive explicit error diagnoses and revision guidance from self-generated attempts, yet existing reflection-based methods often combine it with reference information, rich task feedback, or persistent memory. We introduce ReTeach, a Reflective self-distillation framework that constructs its self-Teacher through multi-round reflection and retry using only self-generated attempts and outcome-level verification. Starting from an unsuccessful student rollout, the teacher alternates explicit reflection with renewed attempts until success or the retry budget is exhausted, without reference answers or solutions, external diagnostic feedback, or cross-example memory. Each failed retry informs subsequent reflection, while successful correction provides outcome-level evidence for the potential utility of the resulting teacher context. An outcome-aware selection and weighting strategy distinguishes initially correct, reflection-corrected, and unresolved examples, assigning separate weights to their category-normalized distillation losses. Through on-policy distillation, the student matches the teacher's context-conditioned token-level predictive distributions at prefixes of its own rollouts, transferring the benefits of iterative correction while retaining single-pass inference. Across six benchmarks spanning mathematical reasoning, science question answering, and tool use, ReTeach improves average accuracy over GRPO by 1.39 percentage points.

2. SSCBench: Evaluating the Evidential Validity of Fault-Injection Tests for Tool-Using LLM Agents

来源：arXiv
日期：2026-10-08
作者/机构：Xincheng He, Wanli Dong, Zhaoqiang Guo, Yan Liu, Lei Xu
主题标签：AI-Agent, arXiv
推荐等级：高
分类：cs.SE

一句话结论

本文研究了工具调用LLM智能体故障注入测试的“证据有效性”问题，提出SSCBench基准，强调评估需报告反证可见的时序条件。

核心内容

制定测量协议，判定反驳观测能否在错误事实首次使用前可见并记录实际执行情况。
相同配置在不同执行中证据条件差异大（44次有反证的采纳中仅17次在首次使用前可见）。
首次出错与后续修正未必同步，且自动轨迹分析难以可靠定位首次错误依赖事件。

方法与数据

构建SSCBench，在2个 $\tau$ -bench环境中评估4种故障算子、5种智能体配置的1,191次故障执行。

价值判断

值得关注：将故障采纳判为失败前，必须报告执行的证据条件及有效样本量。
可复用点：证据有效性测量协议与SSCBench基准。
局限/待核查：自动轨迹分析在时序诊断上不可靠；具体故障算子与配置细节摘要未明确。

摘要

Fault injection is increasingly used to evaluate the reliability of tool-using LLM agents. However, there has been limited study of how fault-adoption results should be interpreted when the agent itself determines which authoritative observations become visible during execution. In this paper, we present a systematic study of this evidential validity problem in agent fault-injection evaluation. We develop a measurement protocol that specifies what observations can refute an injected assertion, determines whether they can become visible before the affected fact is first used, and records whether the evaluated execution actually realizes this condition. We construct SSCBench as an instantiation of the protocol and evaluate four fault operators and five agent configurations over 1,191 faulted executions in two $τ$-bench environments. Our experiments show that an admitted fault case and agent configuration can realize substantially different evidential conditions across executions, and that aggregate adoption can remain well defined even when the population supporting a timely-counterevidence claim is sparse or absent. For example, among 44 adopted runs in which counterevidence eventually became visible, only 17 received it before first use, while 27 received it afterward. We also find that first-error timing and later stance revision need not coincide, and that automated trajectory analysis can recover adoption without reliably recovering the first faulty-reliance event needed for temporal diagnosis. We argue that the evidential condition realized by an execution and the population supporting a claim-specific interpretation are part of fault-injection evaluation itself and should be reported before adoption is interpreted as failure under pre-use counterevidence.