Entity · paper

Resist and Update: Counterfactual Report Coordinates for Incentive-Compatible LLMs

paperactiveprovisionalresist-and-update-counterfactual-report-coordinates-for-incentive-compatible-llms-ddf404b1·1 events·first seen 6d ago

Aliases: Resist and Update: Counterfactual Report Coordinates for Incentive-Compatible LLMs

Co-occurring entities

SycophancyEval Counterfactual Report Coordinates

More like this (12)

Counterfactual Report Coordinates Can LLMs Reliably Self-Report Adversarial Prefills, and How?Janus: A Benchmark for Goal-Conditioned Information Distortion in LLMs RAS: Measuring LLM Safety Through Refusal Alignment Reinforcement Learning with Metacognitive Feedback Elicits Faithful Uncertainty Expression in LLMs Knowledge Knows, Verbalization Tells: Disentangling Latent Directions for Mathematical Solvability in LLMs Which Models Are Our Models Built On? Auditing Invisible Dependencies in Modern LLMs Be My Tutor: On-Policy Co-Distillation for Mutual LLM Improvement via Peer Feedback From Plausible to Actionable: A Position on LLM Self-Explanations Will the Agent Recuse Itself? Measuring LLM-Agent Compliance with In-Band Access-Deny Signals Agentic Chain-of-Thought Steering for Efficient and Controllable LLM Reasoning Accuracy and Satisfaction in Multi-Turn LLM Dialogues for NFR Assessment

Recent events (1)

6arXiv · cs.AI·6d ago·source ↗

Counterfactual Report Coordinates method for certifying incentive-compatible LLM honesty

A new arXiv paper introduces Counterfactual Report Coordinates (CRC), a method for identifying and controlling low-rank activation-space coordinates governing a model's answer, confidence, and caveat outputs, with the goal of making LLMs internally incentive-compatible — resistant to sycophantic pressure while remaining responsive to genuine evidence. The authors use causal interchange interventions (not probe accuracy) to identify these coordinates and introduce a training-free two-pass clamp that achieves near-perfect resist and update scores on a Bayesian-witness benchmark. Results transfer across three model families and to SycophancyEval, though the deployable single-pass compilation is lossy. The work frames activation-level counterfactual incentive-invariance as a structural primitive for alignment research.

Evaluation and Benchmarking AI Safety Research Resist and Update: Counterfactual Report Coordinates for Incentive-Compatible LLMs SycophancyEval Counterfactual Report Coordinates +1 more