Human Interventions as Labels: Localizing, Steering, and (Sometimes) Editing the Computations Behind LLM Agent Failures

Key points

• We treat it instead as a measurement: a label pointing at the internal computation that failed.1
• We study it in a controlled agent environment, on one failure (asking a clarifying question when the request was already unambiguous), with labels from a rule-based overseer validated against blind human judgment.1
• First, the ask-vs-act decision is causally localizable in every eligible family: a direction learned from intervention labels flips 72-100% of held-out over-asks at a definite residual-stream depth, with site and dose chosen on training data, and random directions under the same procedure flip at most one episode.1
• Fourth, consolidating the fix into the weights yields a map, not a rule: on one model a LoRA confined to the causal band passes every criterion fixed in advance while behavior cloning on identical data breaks retention; on three families the reverse holds; on the fifth the outcome crosses the boundary across seeds.1

Sources (1)

• [1]Human Interventions as Labels: Localizing, Steering, and (Sometimes) Editing the Computations Behind LLM Agent Failures
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 15, 11:27 AM

  When an LLM agent misbehaves, a human steps in, and that intervention is usually discarded or replayed as training data.

  We treat it instead as a measurement: a label pointing at the internal computation that failed.

Extractive summary: sentences quoted from the sources.