4 papers
Building Legal Reward Models for Grounding and Abstention
Rilton Franzone, Valentin Noël, Puyu Wang +2
Large language models are increasingly used in high-stakes domains such as law, where systems must ground their reasoning in retrieved evidence and abstain when that evidence is in…
Where You Measure Decides What You Measure: Position Selection in Ablation-Based SAE Evaluation
Valentin Noël
Sparse autoencoders are meant to name the things a language model computes, and the usual way to check that a latent matters is to switch it off and see what changes. But a latent…
A Probe Direction Is a Property of Its Prompt
Valentin Noël
A model that behaves differently when it senses it is being tested would undermine the evaluations we rely on, so recent work has sought to read that sense directly from a model's…
Geometry of Reason: Spectral Signatures of Valid Mathematical Reasoning
Valentin Noël
Verifying whether a language model is genuinely reasoning or pattern-matching remains an open problem: learned verifiers are expensive, and output-based heuristics are brittle. We…