mlsanityData to 14 Oct 2026

Deep Minds and Shallow Probes

arXiv 2605.11448 ↗First posted 12 May 2026cs.LG cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Neural representations are not unique objects. Even when two systems realize the same downstream computation, their hidden coordinates may differ by reparameterization. A probe family intended to reveal structure already present in a representation should therefore be stable under the relevant representation symmetries rather than be tied to a particular basis. We prove that, under standard nondegeneracy assumptions, behaviorally equivalent LLMs have hidden representations related by an invertible affine transformation. The resulting symmetry principle singles out a unique hierarchy of shallow coordinate-stable probes, with linear probes as its degree-1 member. We also show that a natural object for cross-model probe transfer is a shared probe-visible quotient--the representation modulo directions invisible to the probe family--rather than the full hidden state. Experiments on synthetic and real-world tasks support both predictions, showing where degree-2 probes help beyond linear ones and how quotient-based transfer enables coverage-aware monitor portability across model families. These results point toward a broader geometric representation theory of neural probing, with coverage-aware monitor transfer as a concrete operational consequence.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 17 of its 56 references that are in the map. First posted on arXiv 12 May 2026.

Builds on

17 of 56 references.

References · 17

  1. 2024Qwen2.5 Technical Reportcited by 4,942
  2. 2023Mistral 7Bcited by 3,164
  3. 2016Understanding intermediate layers using linear classifier probescited by 1,353
  4. 2023Representation Engineering: A Top-Down Approach to AI Transparencycited by 1,336
  5. 2023Inference-Time Intervention: Eliciting Truthful Answers from a Language Modelcited by 1,302
  6. 2024Refusal in Language Models Is Mediated by a Single Directioncited by 1,032
  7. 2023Steering Language Models With Activation Engineeringcited by 922
  8. 2023The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasetscited by 712
  9. 2023Eliciting Latent Predictions from Transformers with the Tuned Lenscited by 623
  10. 2025Constitutional Classifiers: Defending against Universal Jailbreaks across Thousands of Hours of Red Teamingcited by 181
  11. 2023Large Language Models can Strategically Deceive their Users when Put Under Pressurecited by 146
  12. 2025Auditing language models for hidden objectivescited by 83
5 more
  1. 2023Eliciting Latent Knowledge from Quirky Language Modelscited by 62
  2. 2026Constitutional Classifiers++: Efficient Production-Grade Defenses against Universal Jailbreakscited by 41
  3. 2025Transferring Linear Features Across Language Models With Model Stitchingcited by 21
  4. 2025When Truthful Representations Flip Under Deceptive Instructions?cited by 10
  5. 2024How not to Stitch Representations to Measure Similarity: Task Loss Matching versus Direct Matchingcited by 5

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Loss-Invariant Projections as Passive Probes of Learned Representations
  2. 2025Spectral Identifiability for Interpretable Probe Geometry
  3. 2026Linear Probes Detect Task Format, Not Reasoning Mode in Language Model Hidden States
  4. 2025Into the Rabbit Hull: From Task-Relevant Concepts in DINO to Minkowski Geometry
  5. 2026What Linear Probes Miss: Multi-View Probing for Weight-Space Learning
  6. 2026How Do Linear Probes Emerge? A Circuit-Tracing Framework with Concept-Targeted Attribution
  7. 2026Probing the Latent World: Emergent Discrete Symbols and Physical Structure in Latent Representations
  8. 2026The Topological Dual of a Dataset: A Logic-to-Topology Encoding for AlphaGeometry-Style Data