MENTIS: What Belief Changes Under Alignment? Measuring Multi-Scale Latent Torsion in Language Models

• URL: http://arxiv.org/abs/2606.01060v2
• Date: Sun, 07 Jun 2026 07:03:09 GMT
• Title: MENTIS: What Belief Changes Under Alignment? Measuring Multi-Scale Latent Torsion in Language Models
• Abstract summary: We introduce MENTIS, a framework for measuring alignment-induced internal reorganization.<n>Across four 7-8B model pairs on LITMUS, our study reveals that alignment-induced change is selective rather than uniform.<n>These results suggest preference alignment leaves structured, depth-localized geometric signatures in internal computation beyond what behavior-level evaluation alone can reveal.
• Score: 16.071341117554038
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Preference alignment has substantially improved the observable behavior of large language models, yet it remains unclear what alignment changes internally. Aligned systems still fail under jailbreaks, prompt injection, and retrieval-time corruption, suggesting behavior-level evaluation alone is incomplete. Post-training should leave measurable traces in internal computation. We ask: when an instruction-tuned (IT) model becomes a preference-aligned (PA) model, what geometric structure changes, where do those changes concentrate, and how selectively do they vary across concepts, prompts, and model families? We introduce MENTIS, a geometry-first framework for measuring alignment-induced internal reorganization in paired checkpoints. MENTIS compares IT and PA models using a primary layerwise covariance-based torsion norm (T1), a secondary spectral torsion diagnostic (T2), and an Energy-Radiance-Activation measure (ERA) for depth localization. Across four 7-8B model pairs on LITMUS, our study reveals that alignment-induced change is selective rather than uniform: normative concepts exhibit larger torsion shifts than factual concepts on average; torsion is negatively correlated with contextual entropy; and peak effects localize to architecture-specific mid-to-late layers. The same pattern appears across word-level, prompt-level, and model-level analyses. These results suggest preference alignment leaves structured, depth-localized geometric signatures in internal computation beyond what behavior-level evaluation alone can reveal.

Related papers

• MindTopo: Can Foundation Models Reason in Topological Space? [52.56750995613252]
  We introduce MindTopo, a benchmark of topological intuition across five properties grounded in cognitive science and formal topology.<n>MindTopo evaluates each property at two cognitive levels. Reasoning asks a model to identify topological relations or infer how they change.<n>Every MLLM performs better on reasoning than on planning, and the best-performing model remains far below observed human performance.
  arXiv  Detail & Related papers  (2026-09-10T17:54:32Z)
• The information geometry of large language models is shared, learned, and controllable [0.0]
  We show that behaviour determines geometry up to output-preserving symmetries.<n>Across transformer, state-space and recurrent models, output geometries agree more strongly than activation geometries.<n> Controlled language assignments show that geometry follows the language law across architectures.
  arXiv  Detail & Related papers  (2026-09-10T04:09:49Z)
• Rewriting or Reweighting? A Geometric Account in Language Models [44.66056836026845]
  Post-training can substantially alter language-model behavior.<n>We study two mechanistically distinct failures, repetition as a decoding-attractor pathology and sycophancy as a preference-related alignment failure.<n>We introduce behavioral manifold analysis, which isolates behavior-specific geometry by selecting sparse behavior-associated coordinates.
  arXiv  Detail & Related papers  (2026-08-03T07:46:28Z)
• Geometry-Adaptive Explainer for Faithful Dictionary-Based Interpretability under Distribution Shift [17.611062308867275]
  We show that distribution shift rotates the subspace that the model actively uses, misaligning the explainer's dictionary trained on in-distribution (ID) activations.<n>We formalize this misalignment as the faithfulness gap, a geometric distance between the ID dictionary and the OOD-active subspace.<n>We propose the Geometry-Adaptive Explainer (GAE), which realigns the explainer's dictionary with the OOD-active subspace while preserving the original feature structure.
  arXiv  Detail & Related papers  (2026-05-21T00:46:01Z)
• SpatialEvo: Self-Evolving Spatial Intelligence via Deterministic Geometric Environments [75.60795462502949]
  SpatialEvo is a self-evolving framework for 3D spatial reasoning.<n>It formalizes 16 spatial reasoning task categories under explicit geometric validation rules.<n>It converts unannotated 3D scenes into zero-noise interactive oracles, replacing model consensus with objective physical feedback.
  arXiv  Detail & Related papers  (2026-04-15T17:59:12Z)
• Beyond the Covariance Trap: Unlocking Generalization in Same-Subject Knowledge Editing for Large Language Models [18.348253900450093]
  We show that models fail to recall updated knowledge when following user instructions, despite successfully recalling it in the original edited form.<n>We introduce RoSE (Robust Same-subject Editing), which employs Isotropic Geometric Alignment to minimize representational deviation and Hierarchical Knowledge Integration to smooth the optimization landscape.
  arXiv  Detail & Related papers  (2026-03-16T16:44:54Z)
• Multi-Way Representation Alignment [19.53606443098969]
  Current strategies for mapping representations are inherently pairwise, scaling quadratically with the number of models and failing to yield a consistent global reference.<n>We first adapt Generalized Procrustes Analysis (GPA) to construct a shared universe that preserves the internal geometry essential for tasks like model stitching.<n>We then show that strict isometric alignment is suboptimal for retrieval, where agreement-maximizing methods like Canonical Correlation Analysis (CCA) typically prevail.<n>To bridge this gap, we finally propose Geometry-Corrected Procrustes Alignment (GCPA), which establishes a robust GPA-based universe followed by a post-
  arXiv  Detail & Related papers  (2026-02-05T21:33:45Z)
• Understanding the Implicit Biases of Design Choices for Time Series Foundation Models [90.894232610821]
  Time series foundation models (TSFMs) are a class of potentially powerful, general-purpose tools for time series forecasting and related temporal tasks.<n>Their behavior is strongly shaped by subtle inductive biases in their design.<n>We show how these biases can be intuitive or very counterintuitive, depending on properties of the model and data.
  arXiv  Detail & Related papers  (2025-10-22T04:42:35Z)
• Understanding Post-Training Structural Changes in Large Language Models [3.054513120350576]
  Post-training fundamentally alters the behavior of large language models (LLMs)<n>This work focuses on two widely adopted post-training methods: instruction tuning and long-chain-of-thought (Long-CoT) distillation.
  arXiv  Detail & Related papers  (2025-09-22T15:03:36Z)
• Behavioral Fingerprinting of Large Language Models [35.18856642496912]
  Current benchmarks for Large Language Models (LLMs) primarily focus on performance metrics.<n>This paper introduces a novel Behavioral Fingerprinting'' framework designed to move beyond traditional evaluation.
  arXiv  Detail & Related papers  (2025-09-02T07:03:20Z)
• PAID: Pairwise Angular-Invariant Decomposition for Continual Test-Time Adaptation [70.98107766265636]
  This paper takes the geometric attributes of pre-trained weights as a starting point, systematically analyzing three key components: magnitude, absolute angle, and pairwise angular structure.<n>We find that the pairwise angular structure remains stable across diverse corrupted domains and encodes domain-invariant semantic information, suggesting it should be preserved during adaptation.
  arXiv  Detail & Related papers  (2025-06-03T05:18:15Z)
• Adaptive Spot-Guided Transformer for Consistent Local Feature Matching [64.30749838423922]
  We propose Adaptive Spot-Guided Transformer (ASTR) for local feature matching. ASTR models the local consistency and scale variations in a unified coarse-to-fine architecture.
  arXiv  Detail & Related papers  (2023-03-29T12:28:01Z)
• Robust Change Detection Based on Neural Descriptor Fields [53.111397800478294]
  We develop an object-level online change detection approach that is robust to partially overlapping observations and noisy localization results. By associating objects via shape code similarity and comparing local object-neighbor spatial layout, our proposed approach demonstrates robustness to low observation overlap and localization noises.
  arXiv  Detail & Related papers  (2022-08-01T17:45:36Z)

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.