Anatomy of Post-Training: Using Interpretability to Characterize Data and Shape the Learning Signal

• URL: http://arxiv.org/abs/2606.12360v2
• Date: Thu, 11 Jun 2026 17:45:49 GMT
• Title: Anatomy of Post-Training: Using Interpretability to Characterize Data and Shape the Learning Signal
• Abstract summary: We introduce a data-centric post-training pipeline that uses interpretability protocols to develop statistical hypotheses for the latent concepts separating preferred from dispreferred generations.<n>We show that our pipeline diagnoses undesirable signals in existing preference data, mitigates off-target learning, and can also help amplify or shape desired properties such as safeguards and model personality.
• Score: 29.196846051811576
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Language-model post-training is the main stage at which model behavior is shaped, yet it still largely involves optimization of scalar rewards that summarize diverse desiderata. This abstraction gives practitioners little visibility into what their data actually teaches models, allowing spurious correlations to be learned by a model and inducing undesirable behaviors such as over-stylization and sycophancy. To address this problem, we ask: can we inspect a preference dataset before optimization and decide, at the level of concepts, which behaviors a model should be allowed to learn? Motivated by this, we introduce a data-centric post-training pipeline that uses interpretability protocols to develop statistical hypotheses for the latent concepts separating preferred from dispreferred generations, making them explicit for fine-grained user feedback. Building on this view, we unify several interpretability-based training protocols as ways of shaping rewards via feature or data interventions. Empirically, we show that our pipeline diagnoses undesirable signals in existing preference data, mitigates off-target learning, and can also help amplify or shape desired properties such as safeguards and model personality. More broadly, our results suggest that interpretability can turn post-training from optimizing opaque proxy rewards into a process of auditing and sculpting the learning signal itself.

Related papers

• Supervised learning pays attention [42.97070083645048]
  In-context learning with attention enables large neural networks to make context-specific predictions by selectively focusing on relevant examples.<n>We show how to flexibly fit personalized models for each prediction point and (2) model retain simplicity and interpretability.<n>Our method fits a local model for each test observation by weighting the training data according to attention, a supervised similarity measure.
  arXiv  Detail & Related papers  (2025-12-10T18:43:46Z)
• Did Models Sufficient Learn? Attribution-Guided Training via Subset-Selected Counterfactual Augmentation [61.248535801314375]
  Subset-Selected Counterfactual Augmentation (SS-CA)<n>We develop Counterfactual LIMA to identify minimal spatial region sets whose removal can selectively alter model predictions.<n>Experiments show that SS-CA improves generalization on in-distribution (ID) test data and achieves superior performance on out-of-distribution (OOD) benchmarks.
  arXiv  Detail & Related papers  (2025-11-15T08:39:22Z)
• Revisiting Data Attribution for Influence Functions [13.88866465448849]
  This paper comprehensively reviews the data attribution capability of influence functions in deep learning.<n>We discuss their theoretical foundations, recent algorithmic advances for efficient inverse-Hessian-vector product estimation, and evaluate their effectiveness for data attribution and mislabel detection.
  arXiv  Detail & Related papers  (2025-08-10T11:15:07Z)
• Preference Learning for AI Alignment: a Causal Perspective [55.2480439325792]
  We frame this problem in a causal paradigm, providing the rich toolbox of causality to identify persistent challenges.<n>Inheriting from the literature of causal inference, we identify key assumptions necessary for reliable generalisation.<n>We illustrate failure modes of naive reward models and demonstrate how causally-inspired approaches can improve model robustness.
  arXiv  Detail & Related papers  (2025-06-06T10:45:42Z)
• Capturing the Temporal Dependence of Training Data Influence [100.91355498124527]
  We formalize the concept of trajectory-specific leave-one-out influence, which quantifies the impact of removing a data point during training.<n>We propose data value embedding, a novel technique enabling efficient approximation of trajectory-specific LOO.<n>As data value embedding captures training data ordering, it offers valuable insights into model training dynamics.
  arXiv  Detail & Related papers  (2024-12-12T18:28:55Z)
• Adjusting Pretrained Backbones for Performativity [34.390793811659556]
  We propose a novel technique to adjust pretrained backbones for performativity in a modular way. We show how it leads to smaller loss along the retraining trajectory and enables us to effectively select among candidate models to anticipate performance degradations.
  arXiv  Detail & Related papers  (2024-10-06T14:41:13Z)
• Explanatory Model Monitoring to Understand the Effects of Feature Shifts on Performance [61.06245197347139]
  We propose a novel approach to explain the behavior of a black-box model under feature shifts. We refer to our method that combines concepts from Optimal Transport and Shapley Values as Explanatory Performance Estimation.
  arXiv  Detail & Related papers  (2024-08-24T18:28:19Z)
• Outlier Gradient Analysis: Efficiently Identifying Detrimental Training Samples for Deep Learning Models [24.073260299592675]
  In this paper, we establish a bridge between identifying detrimental training samples via influence functions and outlier gradient detection.<n>We first validate the hypothesis of our proposed outlier gradient analysis approach on synthetic datasets.<n>We then demonstrate its effectiveness in detecting mislabeled samples in vision models and selecting data samples for improving performance of natural language processing transformer models.
  arXiv  Detail & Related papers  (2024-05-06T21:34:46Z)
• Secrets of RLHF in Large Language Models Part II: Reward Modeling [134.97964938009588]
  We introduce a series of novel methods to mitigate the influence of incorrect and ambiguous preferences in the dataset. We also introduce contrastive learning to enhance the ability of reward models to distinguish between chosen and rejected responses.
  arXiv  Detail & Related papers  (2024-01-11T17:56:59Z)
• On the Informativeness of Supervision Signals [31.418827619510036]
  We use information theory to compare how a number of commonly-used supervision signals contribute to representation-learning performance. Our framework provides theoretical justification for using hard labels in the big-data regime, but richer supervision signals for few-shot learning and out-of-distribution generalization.
  arXiv  Detail & Related papers  (2022-11-02T18:02:31Z)
• Distilling Interpretable Models into Human-Readable Code [71.11328360614479]
  Human-readability is an important and desirable standard for machine-learned model interpretability. We propose to train interpretable models using conventional methods, and then distill them into concise, human-readable code. We describe a piecewise-linear curve-fitting algorithm that produces high-quality results efficiently and reliably across a broad range of use cases.
  arXiv  Detail & Related papers  (2021-01-21T01:46:36Z)
• Adversarial Infidelity Learning for Model Interpretation [43.37354056251584]
  We propose a Model-agnostic Effective Efficient Direct (MEED) IFS framework for model interpretation. Our framework mitigates concerns about sanity, shortcuts, model identifiability, and information transmission. Our AIL mechanism can help learn the desired conditional distribution between selected features and targets.
  arXiv  Detail & Related papers  (2020-06-09T16:27:17Z)

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.