Consistency Training while Mitigating Obfuscation via Rate Matching

• URL: http://arxiv.org/abs/2606.02211v1
• Date: Mon, 01 Jun 2026 13:10:49 GMT
• Title: Consistency Training while Mitigating Obfuscation via Rate Matching
• Abstract summary: 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.
• Score: 6.354949885860511
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large language models are often influenced by extraneous input features, such as cues revealing a user's preferred answer. Consistency training reduces this influence by training models to behave similarly across inputs with and without the extraneous feature. However, existing methods train for consistency over entire responses or internal activations, which also constrains whether the model verbalises said extraneous features. We show this leads to obfuscation, where the model learns not to mention a cue while remaining influenced by it, which may undermine monitorability. To address this, we introduce Rate Matching Consistency Training (RMCT), which trains for consistency over selected behavioural properties without constraining how this behaviour is expressed. RMCT matches the rate at which the model exhibits a target behaviour (e.g., following a bias cue) across input perturbations, rather than requiring paired inputs with and without the extraneous feature, extending consistency training to settings where the extraneous features cannot be removed. We evaluate RMCT on sycophancy reduction in two open-weight language models, achieving reductions in bias-following comparable to a standard consistency-training baseline on held-out bias types, while largely preserving the model's tendency to verbalise the bias cue. Further, we find that RMCT is more data-efficient at the expense of being less compute-efficient in our experiments. Overall, RMCT shows that consistency training can improve behavioural robustness without directly trading off against monitorability.

Related papers

• Evaluating the Representation Space of Diffusion Models via Self-Supervised Principles [35.85907484884706]
  We introduce a framework for evaluating the representation and generation capabilities of diffusion models.<n>We analyze both discriminative and generative behavior of diffusion models.<n>Our results show that diffusion models can be monitored from a self-supervised perspective.
  arXiv  Detail & Related papers  (2026-06-08T16:44:18Z)
• Dual-level Modality Debiasing Learning for Unsupervised Visible-Infrared Person Re-Identification [59.59359638389348]
  We propose a Dual-level Modality Debiasing Learning framework that implements debiasing at both the model and optimization levels.<n>Experiments on benchmark datasets demonstrate that DMDL could enable modality-invariant feature learning and a more generalized model.
  arXiv  Detail & Related papers  (2025-12-03T12:43:16Z)
• Ascent Fails to Forget [45.75497227694833]
  We show that gradient ascent-based unconstrained optimization methods frequently fail to perform machine unlearning.<n>We attribute this phenomenon to the inherent statistical dependence between the forget and retain data sets.<n>Our findings highlight that the presence of such statistical dependencies, even when manifest only as correlations, can be sufficient for ascent-based unlearning to fail.
  arXiv  Detail & Related papers  (2025-09-30T15:48:49Z)
• Dissecting Representation Misalignment in Contrastive Learning via Influence Function [15.28417468377201]
  We introduce the Extended Influence Function for Contrastive Loss (ECIF), an influence function crafted for contrastive loss.<n>ECIF considers both positive and negative samples and provides a closed-form approximation of contrastive learning models.<n>Building upon ECIF, we develop a series of algorithms for data evaluation, misalignment detection, and misprediction trace-back tasks.
  arXiv  Detail & Related papers  (2024-11-18T15:45:41Z)
• Noisy Correspondence Learning with Self-Reinforcing Errors Mitigation [63.180725016463974]
  Cross-modal retrieval relies on well-matched large-scale datasets that are laborious in practice. We introduce a novel noisy correspondence learning framework, namely textbfSelf-textbfReinforcing textbfErrors textbfMitigation (SREM)
  arXiv  Detail & Related papers  (2023-12-27T09:03:43Z)
• Mitigating Shortcut Learning with Diffusion Counterfactuals and Diverse Ensembles [104.60508550106618]
  We propose DiffDiv, an ensemble diversification framework exploiting Diffusion Probabilistic Models (DPMs)<n>We show that DPMs can generate images with novel feature combinations, even when trained on samples displaying correlated input features.<n>We show that DPM-guided diversification is sufficient to remove dependence on shortcut cues, without a need for additional supervised signals.
  arXiv  Detail & Related papers  (2023-11-23T15:47:33Z)
• Explicit Tradeoffs between Adversarial and Natural Distributional Robustness [48.44639585732391]
  In practice, models need to enjoy both types of robustness to ensure reliability. In this work, we show that in fact, explicit tradeoffs exist between adversarial and natural distributional robustness.
  arXiv  Detail & Related papers  (2022-09-15T19:58:01Z)
• Agree to Disagree: Diversity through Disagreement for Better Transferability [54.308327969778155]
  We propose D-BAT (Diversity-By-disAgreement Training), which enforces agreement among the models on the training data. We show how D-BAT naturally emerges from the notion of generalized discrepancy.
  arXiv  Detail & Related papers  (2022-02-09T12:03:02Z)
• Understanding the Logit Distributions of Adversarially-Trained Deep Neural Networks [6.439477789066243]
  Adversarial defenses train deep neural networks to be invariant to the input perturbations from adversarial attacks. Although adversarial training is successful at mitigating adversarial attacks, the behavioral differences between adversarially-trained (AT) models and standard models are still poorly understood. We identify three logit characteristics essential to learning adversarial robustness.
  arXiv  Detail & Related papers  (2021-08-26T19:09:15Z)
• Robust Pre-Training by Adversarial Contrastive Learning [120.33706897927391]
  Recent work has shown that, when integrated with adversarial training, self-supervised pre-training can lead to state-of-the-art robustness. We improve robustness-aware self-supervised pre-training by learning representations consistent under both data augmentations and adversarial perturbations.
  arXiv  Detail & Related papers  (2020-10-26T04:44:43Z)

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.