Drop the Act: Probe-Filtered RL for Faithful Chain-of-Thought Reasoning

TL;DR

ProFIL is a reinforcement learning extension that uses a probe to detect and suppress unfaithful reasoning steps in models, improving interpretability and efficiency without sacrificing accuracy.

Contribution

It introduces a probe-based method to reduce reasoning theater in models, enhancing faithfulness and chain efficiency across multiple domains.

Findings

ProFIL reduces reasoning theater by 11-100% across tasks.

It increases the faithfulness of model reasoning by up to 24 percentage points.

Chains are shortened by 4-19% while maintaining or improving accuracy.

Abstract

Reasoning models post-hoc rationalize answers they have already committed to internally, producing chains of *reasoning theater*: deliberative-looking steps that contribute nothing to correctness. This wastes inference tokens, pollutes interpretability, and obscures what the model actually computed. We introduce **ProFIL** (**Pro**be-**Fil**tered Reinforcement Learning) to *reduce theater, increase chain-of-thought faithfulness, and shrink chain length* in a single, drop-in extension to Group Relative Policy Optimization (GRPO). A multi-head attention probe is trained *once* on the *frozen* base model to detect post-commitment steps from internal activations alone; during GRPO, rollouts whose probe score exceeds a threshold have their advantage zeroed. *Our central finding is that a probe trained on a frozen base, with verifier-derived labels and no human annotation, provides a stable signal that suppresses theater while resisting the RL-obfuscation failure mode predicted by prior work.* Across four reasoning domains (GSM8K, LiveCodeBench, ToolUse, MMLU-Redux) and two model architectures (Llama-8B, Qwen-7B), ProFIL reduces post-commitment theater by **11--100%**, raises faithful-fraction (e.g., +24pp on LiveCodeBench under an independent Claude 3.7 Sonnet judge), and shortens chains by 4--19%, all while preserving or improving task accuracy. ProFIL also beats a matched length-penalty GRPO baseline, isolating the gain as semantic commitment-detection rather than chain compression. Probe weights, training configurations, and rollouts are released across all four domains.