Skip to content

Alignment, Reasoning, and Trustworthiness

LP

In brief — site editorial, not from the paper

Trains a small classifier on the model's internal representations to see what information they contain, without changing the model itself.

Definition

LP is framed as an interpretability and explainability intervention—determining its D2 (Goal) assignment as Explainability—rather than merely as a downstream evaluation metric, task performance, or behavioral alignment [222]. By freezing all base model parameters and training only a lightweight linear classifier appended to a designated intermediate layer, LP surfaces latent representational structure as interpretable task-specific outputs without modifying the base model's core feature extraction function [16]. The Parametric Update classification applies exclusively to the probe head's weight matrix; the base remains mathematically unchanged. Its Modular scope indicates that LP probes constitute structurally decoupled artifacts whose installation and removal leave the base model's parametric state and validated behavior unaffected. Compared to PEFT, which also introduces a modular artifact onto a frozen base, LP diverges critically on D2 (Goal): PEFT targets task performance (Computational Efficiency, Task Specialization), while LP targets representational transparency. Compared to LEACE, which operates via Parameter Composition to algebraically rewrite the base model's weight matrices, LP never touches the base weights at all. For production deployments requiring verifiable transparency, the probe's classification performance across sensitive concept dimensions can be documented as engineering evidence of representational alignment or bias. Because LP does not modify the base model's parameters, it does not trigger a model-level change-control event; the decoupled probe artifact (Modular) requires its own version-controlled documentation specifying the target layer, concept labels, and labeled probe dataset—but the base model's existing validation record remains entirely unaffected.

Verbatim from the paper — Alignment, Reasoning, and Trustworthiness

Notes from the table

Classification tensions involving these techniques are discussed in Supplementary Section S4.

Classification tensions

LP. A strictly frozen linear probe does not modify the base model's internal representations or predictive behavior; it merely surfaces intermediate states. It is included in the adaptation framework strictly because appending a probe fundamentally alters the deployed system's explainability footprint and output characteristics (Explainability), triggering regulatory transparency mandates (e.g., EU AI Act Article 13) even when the core parameters remain untouched.

Appendix C

Nearest profiles

PEFT (LoRA, adapters) (0.25), ReFT (0.25), Calibration (0.28), FT (partial) (0.33), Prompt Learning (0.38)

Computed from the taxonomy data (Gower distance over all six dimensions)

References

  1. [16] Guillaume Alain, and Yoshua Bengio 2018. Understanding intermediate layers using linear classifier probes. arXiv
  2. [222] Jason Yosinski, Jeff Clune, Yoshua Bengio, and Hod Lipson 2014. How transferable are features in deep neural networks?. Advances in Neural Information Processing Systems

Numbered as in the paper

Cite this row

LP — six-dimensional profile
D1: parametric-update
D2: explainability
D3: small-labeled
D4: ad-hoc-permanent, scheduled-permanent
D5: modular
D6: dl, fm, llm, mllm
Source: arXiv:2608.06246
Compare with another technique Find in the explorer