Research

1 item

arXiv5 Nov 2026Interpretability, Control

Failure Prediction and Classification in Vision-Language-Action Models Via Calibrated Latent Probing

Michal P. Podolinsky, Yunhao Yang, Daniel Little, Hong Lu, Rohan Siva and 4 more

Proposes VLA-Tug, a lightweight calibrated probe on VLA latent layers that gives runtime failure detection with bounded error rates, predicts failure modes, and enables steering under distribution shift.

2 items

LessWrong4 Nov 2026Interpretability, Control

Model's "emotions" noticeably change when they take risky actions

jain18ayush, ShivamR, conti, sophia_xu and johnyan

Shows that emotion-vector probes in LLM activations shift before reward hacking and specification gaming, can predict hacks several sentences early, and steering them changes hacking rates.

arXiv4 Nov 2026Interpretability, Control

Shaping & Monitoring Models through their Internal Representations

Thomas Fel, Matthew Kowal, Matt Feiszli, Satchel Grant, Andrew Lee and 21 more

Proposes Mechanistic Shaping, training models against internal representations to suppress unsafe concepts, and studies whether shaped concepts stay monitorable via probes, finding three circumvention mechanisms.

Research

Filters

Focus

Date

Sort

Failure Prediction and Classification in Vision-Language-Action Models Via Calibrated Latent Probing

Model's "emotions" noticeably change when they take risky actions

Shaping & Monitoring Models through their Internal Representations

Research

Thursday 5 November

Failure Prediction and Classification in Vision-Language-Action Models Via Calibrated Latent Probing

Wednesday 4 November

Model's "emotions" noticeably change when they take risky actions

Shaping & Monitoring Models through their Internal Representations