Towards interpreting any residual stream direction by scaling activation inversion
cs.LG, cs.AI
Submitted: 2026-11-05
Updated: 2026-11-05
Code: https://github.com/ceselder/maem
Terminology
Sources
- Understanding intermediate layers using linear classifier probes
- Llemma: An Open Language Model For Mathematics
- Building Better Activation Oracles
- BatchTopK Sparse Autoencoders
- SelfIE: Self-Interpretation of Large Language Model Embeddings
- Meta-Models: An Architecture for Decoding LLM Behaviors Through Interpreted Embeddings and Natural Language
- The Vendi Score: A Diversity Evaluation Metric for Machine Learning
- Scaling and evaluating sparse autoencoders
- Patchscopes: A Unifying Framework for Inspecting Hidden Representations of Language Models
- Learning to Interpret Weight Differences in Language Models
- Verbalizable Representations Form a Global Workspace in Language Models
- SAEExplainer: Interpreting SAE Features with Activation-Guided Preference Optimization
- Predictive Concept Decoders: Training Scalable End-to-End Interpretability Assistants
- InversionView: A General-Purpose Method for Reading Information from Neural Activations
- Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers
- The Art of Scaling Reinforcement Learning Compute for LLMs
- The Stack: 3 TB of permissively licensed source code
- Training Language Models to Explain Their Own Computations
- Do Activation Verbalization Methods Convey Privileged Information?
- Eliciting Language Model Behaviors with Investigator Agents
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