The Geometric Canary: Predicting Steerability and Detecting Drift via Representational Stability

• URL: http://arxiv.org/abs/2604.17698v2
• Date: Sat, 25 Apr 2026 22:45:11 GMT
• Title: The Geometric Canary: Predicting Steerability and Detecting Drift via Representational Stability
• Abstract summary: We show that geometric stability, the consistency of a representation's pairwise distance structure, addresses two functions.<n> unsupervised stability fails entirely for steering on real-world tasks.<n>It excels at drift detection, measuring nearly $2times$ greater geometric change than CKA during post-training alignment.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Reliable deployment of language models requires two capabilities that appear distinct but share a common geometric foundation: predicting whether a model will accept targeted behavioral control, and detecting when its internal structure degrades. We show that geometric stability, the consistency of a representation's pairwise distance structure, addresses both. Supervised Shesha variants that measure task-aligned geometric stability predict linear steerability with near-perfect accuracy ($ρ= 0.89$-$0.97$) across 35-69 embedding models and three NLP tasks, capturing unique variance beyond class separability (partial $ρ= 0.62$-$0.76$). A critical dissociation emerges: unsupervised stability fails entirely for steering on real-world tasks ($ρ\approx 0.10$), revealing that task alignment is essential for controllability prediction. However, unsupervised stability excels at drift detection, measuring nearly $2\times$ greater geometric change than CKA during post-training alignment (up to $5.23\times$ in Llama) while providing earlier warning in 73\% of models and maintaining a $6\times$ lower false alarm rate than Procrustes. Together, supervised and unsupervised stability form complementary diagnostics for the LLM deployment lifecycle: one for pre-deployment controllability assessment, the other for post-deployment monitoring.

Related papers

• First-Order Predictable but Pairwise Fragile: Local Task Adaptation in Trained Transformers [0.05729426778193398]
  We measure 8 properties with the same harness around a multitask LoRA operating point.<n>We find a shared one-direction validity window up to the tested scale $10-2$, but no universal radius for pairwise composition or update ordering.
  arXiv  Detail & Related papers  (2026-07-18T13:43:44Z)
• Ask the World Before Acting: Environment Probing for Calibrated Agent World Models [2.571318992830046]
  Language agents acting over long horizons must maintain beliefs about tool states, object locations, graph edges, and subgoal dependencies.<n>We formulate environment probing as a budgeted decision problem for structured agent world models.<n>We introduce EnvProbe, a simple scoring policy that combines task criticality, staleness, verbalized uncertainty, and dependency role.
  arXiv  Detail & Related papers  (2026-06-30T09:45:49Z)
• Conformal Orbit-Valid Trust Horizons for Equivariant World Models [6.230579198456525]
  We study trust-horizon certification for latent world models with known group symmetries.<n>Across 50 stable audits, we observe zero anti-conservative violations.
  arXiv  Detail & Related papers  (2026-06-23T06:44:27Z)
• CIAware-Bench: Benchmarking Control Intervention Awareness Across Frontier LLMs [100.38986535324284]
  We introduce textbfCIAware-Bench, a benchmark for measuring textbfcontrol textbfintervention (CI) awareness across frontier models.<n>CIAware-Bench tests whether models can distinguish their own trajectories from those modified by a control intervention.
  arXiv  Detail & Related papers  (2026-06-09T16:24:16Z)
• CaliDist: Calibrating Large Language Models via Behavioral Robustness to Distraction [51.56484100374058]
  Existing calibration methods for Large Language Models (LLMs) often overlook a critical dimension of trustworthiness: a model's behavioral robustness.<n>We introduce textscCaliDist, a novel post-hoc calibration approach that directly measures and penalizes a model's susceptibility to distraction.<n>textscCaliDist consistently achieves lower Expected Error (ECE) and Brier Score compared with strong baselines.
  arXiv  Detail & Related papers  (2026-06-04T07:27:53Z)
• Representation Without Reward: A JEPA Audit for LLM Fine-Tuning [1.2691047660244335]
  Joint-embedding predictive architectures (JEPAs) propose that a model should learn more useful abstractions when trained to predict latent representations rather than observed outputs.<n>For autoregressive language-model fine-tuning the principle entails a stricter requirement: the induced hidden-state geometry must reach the language-model head emphand improve the decoded task metric.<n>We test that requirement under a fixed Llama-3.2-1B-Instruct LoRA harness on natural-language-to-regex generation.
  arXiv  Detail & Related papers  (2026-05-14T20:27:32Z)
• Tight Convergence Rates for Online Distributed Linear Estimation with Adversarial Measurements [66.94250413799232]
  We study mean estimation of a random vector $X$ in a distributed parameter-server-worker setup.<n>Key challenges are adversarial measurements and asynchrony.<n>Results provide a unified finite-time characterization of robustness, identifiability, and statistical efficiency in distributed linear estimation with adversarial workers.
  arXiv  Detail & Related papers  (2026-04-07T11:45:55Z)
• The Geometry of Robustness: Optimizing Loss Landscape Curvature and Feature Manifold Alignment for Robust Finetuning of Vision-Language Models [33.10473765967503]
  Generalization-preserving methods retain ID/OOD performance but leave models vulnerable to adversarial attacks.<n>Our key insight is that the robustness trade-off stems from two geometric failures: sharp, anisotropic minima in parameter space and unstable feature representations that deform under perturbation.<n>We propose GRACE, a unified fine-tuning framework that jointly regularizes the parameter-space curvature and feature-space invariance for VLMs.
  arXiv  Detail & Related papers  (2026-03-28T05:22:00Z)
• Robustness, Cost, and Attack-Surface Concentration in Phishing Detection [0.0]
  Phishing detectors built on engineered website features attain near-perfect accuracy under i.i.d. evaluation.<n>We study this gap through a cost-aware evasion framework that models discrete, monotone feature edits under explicit attacker budgets.
  arXiv  Detail & Related papers  (2026-03-19T17:53:32Z)
• $V_{0.5}$: Generalist Value Model as a Prior for Sparse RL Rollouts [81.48669089692189]
  Generalist Value Models (such as $V_0.5$) achieve pre-trained value estimation by explicitly encoding model capabilities in-context.<n>We propose $V_0.5$, which adaptively fuses the baseline predicted by such value model with the empirical mean derived from sparse rollouts.<n>$V_0.5$ significantly outperforms GRPO and DAPO, achieving faster convergence and over some 10% performance improvement.
  arXiv  Detail & Related papers  (2026-03-11T14:57:41Z)
• Information Fidelity in Tool-Using LLM Agents: A Martingale Analysis of the Model Context Protocol [69.11739400975445]
  We introduce the first theoretical framework for analyzing error accumulation in Model Context Protocol (MCP) agents.<n>We show that cumulative distortion exhibits linear growth and high-probability deviations bounded by $O(sqrtT)$.<n>Key findings include: semantic weighting reduces distortion by 80%, and periodic re-grounding approximately every 9 steps suffices for error control.
  arXiv  Detail & Related papers  (2026-02-10T21:08:53Z)
• Geometric Stability: The Missing Axis of Representations [0.0]
  We introduce $geometric$ $stability, a distinct dimension that quantifies how reliably representational geometry holds under perturbation.<n>Across 2,463 configurations in seven domains, we show that stability and similarity are empirically uncorrelated.<n>For safety monitoring, stability acts as a functional geometric canary, detecting structural drift nearly 2$times$ more sensitively than CKA.<n>For model selection, stability dissociates from transferability, revealing a geometric tax that transfer optimization incurs.
  arXiv  Detail & Related papers  (2026-01-14T05:15:22Z)
• Adaptive Dual Uncertainty Optimization: Boosting Monocular 3D Object Detection under Test-Time Shifts [80.32933059529135]
  Test-Time Adaptation (TTA) methods have emerged to adapt to target distributions during inference.<n>We propose Dual Uncertainty Optimization (DUO), the first TTA framework designed to jointly minimize both uncertainties for robust M3OD.<n>In parallel, we design a semantic-aware normal field constraint that preserves geometric coherence in regions with clear semantic cues.
  arXiv  Detail & Related papers  (2025-08-28T07:09:21Z)
• Towards Stable 3D Object Detection [64.49059005467817]
  Stability Index (SI) is a new metric that can comprehensively evaluate the stability of 3D detectors in terms of confidence, box localization, extent, and heading. To help models improve their stability, we introduce a general and effective training strategy, called Prediction Consistency Learning (PCL) PCL essentially encourages the prediction consistency of the same objects under different timestamps and augmentations, leading to enhanced detection stability.
  arXiv  Detail & Related papers  (2024-07-05T07:17:58Z)
• Robust Online Control with Model Misspecification [96.23493624553998]
  We study online control of an unknown nonlinear dynamical system with model misspecification. Our study focuses on robustness, which measures how much deviation from the assumed linear approximation can be tolerated.
  arXiv  Detail & Related papers  (2021-07-16T07:04:35Z)

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.