Introspective Coupling: Self-Explanation Training Tracks Behavioral Change Despite Fixed Supervision

• URL: http://arxiv.org/abs/2606.32038v1
• Date: Tue, 30 Jun 2026 17:59:32 GMT
• Title: Introspective Coupling: Self-Explanation Training Tracks Behavioral Change Despite Fixed Supervision
• Abstract summary: We show that LMs trained on fixed counterfactual explanations produce explanations more faithful to their own current behaviors than to those of their training targets.<n>This "introspective" coupling occurs when training explanations remain sufficiently correlated with current behaviors over the course of training.<n>We also show that introspective coupling tracks behavior shifts: when explanation training is provided concurrently with other post-training objectives, explanations track those shifts without requiring updated supervision.
• Score: 49.12474373883331
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: When does training language models (LMs) to generate explanations of their predictions yield faithful introspection, rather than superficial imitation? We study LMs trained to explain which features of their inputs influenced their behavior, using models' counterfactual behavior on modified inputs as supervision. Surprisingly, we find that LMs trained on fixed counterfactual explanations derived from earlier checkpoints of themselves, or even from behaviorally similar models in different families, frequently produce explanations more faithful to their own current behaviors than to those of their training targets. This "introspective" coupling between LM explanations and behaviors occurs when training explanations remain sufficiently correlated with current behaviors over the course of training, even as behaviors themselves shift. We also show that introspective coupling tracks behavior shifts: when explanation training is provided concurrently with other post-training objectives, explanations track those shifts without requiring updated supervision. This phenomenon appears in multiple tasks, including sycophancy and refusal, and is robust to label noise. Overall, our results show that even fixed datasets of counterfactual explanations can provide scalable and generalizable post-training signal for introspection.

Related papers

• Would this change your answer? Evaluating Explanations of LLM Behavior In The Wild with Counterfactual Experiments [4.8434403033301185]
  We introduce CHIVE, a novel agentic pipeline that identifies unexpected model behaviors in the wild and investigates them with counterfactual prompt edits.<n>This yields thousands of high-quality explanations for naturally-occurring model behaviors along with supporting counterfactual evidence.<n>We evaluate whether common LLM interpretability techniques improve an agent's ability to predict counterfactual model behaviors. Surprisingly, we find no uplift from any of the interpretability techniques studied.
  arXiv  Detail & Related papers  (2026-08-17T15:57:06Z)
• Self-CTRL: Self-Consistency Training with Reinforcement Learning [45.9704217495657]
  Language models (LMs) that faithfully describe their own behavior can more easily be audited, understood, and trusted by users.<n>This paper describes Self-Consistency Training with Reinforcement Learning (Self-CTRL), a method that optimize for consistency between a LM's self-explanations and behavior on related inputs.
  arXiv  Detail & Related papers  (2026-06-16T17:59:40Z)
• Forecasting Future Behavior as a Learning Task [25.073460373115733]
  We train Behavior Forecasters that operate on a single reasoning trajectory to make the same forecasts one would typically seek from an explanation.<n>We evaluate this approach on two tasks: how likely the LRM is to repeat its answer on re-runs, and how removing parts of the input changes its answer.
  arXiv  Detail & Related papers  (2026-06-09T20:56:23Z)
• Consistency Training while Mitigating Obfuscation via Rate Matching [6.354949885860511]
  Large language models are often influenced by extraneous input features, such as cues revealing a user's preferred answer.<n>Existing methods train for consistency over entire responses or internal activations, which also constrains whether the model verbalises said extraneous features.<n>We show this leads to obfuscation, where the model learns not to mention a cue while remaining influenced by it, which may undermine monitorability.<n>We introduce Rate Matching Consistency Training (RMCT), which trains for consistency over selected behavioural properties without constraining how this behaviour is expressed.
  arXiv  Detail & Related papers  (2026-06-01T13:10:49Z)
• Revealing Behavioral Plasticity in Large Language Models: A Token-Conditional Perspective [63.9266662672413]
  We show that Large Language Models (LLMs) possess intrinsic behavioral plasticity that can be exposed through token-conditional generation and stabilized via reinforcement learning.<n>We propose Token-Conditioned Reinforcement Learning (ToCoRL) to transform transient inference-time adaptations into stable and learnable behavioral patterns.
  arXiv  Detail & Related papers  (2026-03-09T13:56:53Z)
• Emergently Misaligned Language Models Show Behavioral Self-Awareness That Shifts With Subsequent Realignment [0.3823356975862005]
  We fine-tune GPT-4.1 models sequentially on datasets known to induce and reverse emergent misalignment.<n>Our results show that emergently misaligned models rate themselves as significantly more harmful compared to their base model and realigned counterparts.<n>Our findings show that behavioral self-awareness tracks actual alignment states of models, indicating that models can be queried for informative signals about their own safety.
  arXiv  Detail & Related papers  (2026-02-16T14:29:46Z)
• On the Paradoxical Interference between Instruction-Following and Task Solving [50.75960598434753]
  Instruction following aims to align Large Language Models (LLMs) with human intent by specifying explicit constraints on how tasks should be performed.<n>We reveal a counterintuitive phenomenon: instruction following can paradoxically interfere with LLMs' task-solving capability.<n>We propose a metric, SUSTAINSCORE, to quantify the interference of instruction following with task solving.
  arXiv  Detail & Related papers  (2026-01-29T17:48:56Z)
• The Personality Illusion: Revealing Dissociation Between Self-Reports & Behavior in LLMs [60.15472325639723]
  Personality traits have long been studied as predictors of human behavior.<n>Recent advances in Large Language Models (LLMs) suggest similar patterns may emerge in artificial systems.
  arXiv  Detail & Related papers  (2025-09-03T21:27:10Z)
• CogCoM: A Visual Language Model with Chain-of-Manipulations Reasoning [61.21923643289266]
  Chain of Manipulations is a mechanism that enables Vision-Language Models to solve problems step-by-step with evidence.<n>After training, models can solve various visual problems by eliciting intrinsic manipulations (e.g., grounding, zoom in) actively without involving external tools.<n>Our trained model, textbfCogCoM, achieves state-of-the-art performance across 9 benchmarks from 4 categories.
  arXiv  Detail & Related papers  (2024-02-06T18:43:48Z)
• Explain, Edit, and Understand: Rethinking User Study Design for Evaluating Model Explanations [97.91630330328815]
  We conduct a crowdsourcing study, where participants interact with deception detection models that have been trained to distinguish between genuine and fake hotel reviews. We observe that for a linear bag-of-words model, participants with access to the feature coefficients during training are able to cause a larger reduction in model confidence in the testing phase when compared to the no-explanation control.
  arXiv  Detail & Related papers  (2021-12-17T18:29:56Z)
• Shaking the foundations: delusions in sequence models for interaction and control [45.34593341136043]
  We show that sequence models "lack the understanding of the cause and effect of their actions" leading them to draw incorrect inferences due to auto-suggestive delusions. We show that in supervised learning, one can teach a system to condition or intervene on data by training with factual and counterfactual error signals respectively.
  arXiv  Detail & Related papers  (2021-10-20T23:31:05Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.