Shaping & Monitoring Models through their Internal Representations
cs.LG, cs.AI
Submitted: 2026-11-04
Updated: 2026-11-04
Code: https://github.com/goodfire-ai/reliance
Terminology
Sources
- Subliminal Learning: Language models transmit behavioral traits via hidden signals in data
- Decomposing Query-Key Feature Interactions Using Contrastive Covariances
- Tensor Product Representation Probes Reveal Shared Structure Across Linear Directions
- Features as Rewards: Scalable Supervision for Open-Ended Tasks via Interpretability
- Probe-based Fine-tuning for Reducing Toxicity
- Agentic Reinforcement Learning for Search Misaligns Instruction-Tuning
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis