[ICLR 2025] General-purpose activation steering library
KV Cache Steering for Controlling Frozen LLMs
OpenAI-compatible server with a live view into any HF model's residual stream. pip install brainscope
Agents that talk through model internals — activations & J-space — instead of text. No words pass between them. The lab on top of brainscope + hidden-directions.
[ACL 2026] - Official repo for the paper: "Selective Steering: Norm-Preserving Control Through Discriminative Layer Selection"
[Under Review] Not All Tokens Are Equally Useful for Steering: Robust Directions and Prefix Steering
Activation steering and trait monitoring for HuggingFace transformers
A bilingual awesome list for refusal suppression research: benchmarks, papers, tools, models, and ecosystem updates.
[ICLR 2026] ASGuard: Activation-Scaling Guard to Mitigate Targeted Jailbreaking Attack
Official code for "Activation Steering for Accent Adaptation in Speech Foundation Models" (Interspeech 2026). Parameter-free accent adaptation via mean-shift steering vectors — no weight updates, consistent WER reductions across 8 accents.
Transformer interpretability research framework for circuit discovery, sparse features, and controlled residual-stream interventions.
Runtime rank-1 refusal projection for DeepSeek-V4-Flash-0731: 757KB of directions instead of a 1.54GB weight overlay, lambda as a hot-swappable dial. Full A/B measurements on 2x DGX Spark.
GEMS: Geometric Constraints Enable Multi-Semantic Superposition in LLMs
🏆[ICML 2026 Spotlight] Official implementation of "DecodeShare: Tracing the Shared Subspace of LLM Decode-Time Decisions"
Feature steering for open LLMs: find an interpretable SAE feature, steer the model, measure the causal effect
MiniMax-H3 custom nodes: steering / cache / prompt director
Does concept-injection introspection emerge with scale? A faithful, controlled reproduction charted across model-size ladders.
The paper list related to activation steering
Inspect, steer & monitor a real LLM on Apple Silicon — a browser-based SAE interpretability lab for Qwen Scope, powered by MLX
To associate your repository with the activation-steering topic, visit your repo's landing page and select "manage topics."