今日速览
| 序号 | 标题 | 来源 | 日期 | 主题 | 推荐等级 |
|---|---|---|---|---|---|
| 1 | The Regression Tax: Decomposing Why Skills Help and Hurt LLM Agents | arXiv | 2026-07-24 | RAG | 高 |
| 2 | Where FactsGo Missing: A LayerwiseTaxonomy and Per-Layer Attribution of Information Omissionin Air-Gapped LLM Agent Pipelines | arXiv | 2026-07-24 | AI-Agent | 高 |
重点论文与技术动态
1. The Regression Tax: Decomposing Why Skills Help and Hurt LLM Agents
- 来源：arXiv
- 日期：2026-07-24
- 作者/机构：Darshan Tank, Baran Nama
- 主题标签：
RAG,arXiv - 推荐等级：高
- 分类：cs.AI
一句话结论
评估LLM代理的程序化技能时，应分解净效果（收益与回归）而非仅看平均改进。
一段话：本研究通过近6000次实验（覆盖两个办公室自动化基准和三个模型测试堆栈）对比有无技能的代理，发现技能添加常引发回归（任务从成功变失败）和残留失败（始终失败）。最佳技能表现主要源于减少回归而非增加收益，根源包括技能描述渗透（技能存在即改变行为）、接地位移（覆盖输入解释）和验证位移（抑制输出检查）。分析显示现有技能过度强调程序指导，而忽视接地与验证（主要错误来源），改进后者可恢复多数问题，可靠性更依赖于接地与验证而非技能选择。
核心内容
- 回归（技能添加后任务失败）和残留失败（技能添加前后都失败）是关键区分指标。
- 最佳技能表现主要通过减少回归实现优势，而非增加收益。
- 三种回归机制：技能描述渗透、接地位移、验证位移。
方法与数据
- 实验覆盖近6000次运行，涉及两个办公室自动化基准和三个模型测试堆栈。
价值判断
- 值得关注：技能添加可能使代理性能下降，需评估净效果而非仅平均改进。
- 可复用点：优化接地与验证机制可恢复多数回归和残留失败，提升可靠性。
- 局限/待核查：摘要未明确。
摘要
Adding procedural skills to an LLM agent is typically evaluated by average improvement in task success. However, this metric hides an important cost: skills can also make agents worse. We measure both sides by comparing agents with and without skills across nearly 6,000 runs spanning two office automation benchmarks and three model harness stacks. This allows us to distinguish two outcomes. A regression is a task solved without skills but failed after skills are added. A residual failure is a task that fails both with and without skills. We find that regressions are substantial enough that the best performing skills outperform others primarily by regressing less, not by gaining more. We identify three causes of regression: (i) skill description osmosis, a skill changes an agent's behavior simply by being present in context, even when it is never invoked; (ii) grounding displacement, a skill's prescribed procedure overrides how the agent interprets its inputs; and (iii) verification displacement, where the procedure suppresses checks the agent would otherwise perform on its outputs. Analysing persistent failures reveals the same underlying pattern. Existing skills overemphasize procedural guidance the stage least often responsible for failure while under supporting grounding and verification, the dominant sources of remaining errors. After correcting evaluation artifacts and studying traces, we find many regressions and persistent failures recoverable through better grounding and verification. Procedural skills should be evaluated by decomposing their net effect into gains and regressions, not by aggregate improvement alone. We identify three regression modes skills should avoid, and find that reliability depends more on grounding and verification than on procedural skill choice.2. Where FactsGo Missing: A LayerwiseTaxonomy and Per-Layer Attribution of Information Omissionin Air-Gapped LLM Agent Pipelines
- 来源：arXiv
- 日期：2026-07-24
- 作者/机构：Santhiya Rajan
- 主题标签：
AI-Agent,arXiv - 推荐等级：高
- 分类：cs.MA
一句话结论
信息遗漏在air-gapped LLM代理管道中是主要可靠性故障，68%源于确定性中间件层（L0-L3）。
核心内容
- 提出九层分类法（L0-L8）定位信息遗漏机制，覆盖从数据摄入到代理循环的全过程。
- 开发归因方法，通过受控消融和logit分解分离确定性层（L0-L3）与行为层（L4-L8），量化遗漏率。
- 75,476次试验验证，68%遗漏来自确定性中间件，指导操作者在L0-L3层干预。
方法与数据
- 方法：受控消融和logit分解进行归因。
- 数据：75,476次试验，五种量化4-8B模型（如llama.cpp/vLLM），两种引擎。
价值判断
- 值得关注：临床FHIR服务、法律审查等受监管场景中，信息遗漏导致决策风险。
- 可复用点：九层分类法和归因框架可推广至其他air-gapped系统。
- 局限/待核查：服务器端配置因素（权重量化、KV缓存类型、RoPE缩放）固定，需未来工作。