Steering vectors for transformer language models in Pytorch / Huggingface
-
Updated
Feb 21, 2025 - Python
#
representation-engineering
Here are 48 public repositories matching this topic...
SOM network modified in order to control the latent space
som representation-learning som-clustering manifold-learning latent-space manifold-alignment representation-engineering
-
Updated
May 26, 2025 - Python
[ICLR 2025] General-purpose activation steering library
-
Updated
Sep 18, 2025 - Python
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.
speech-recognition whisper asr interspeech accent-adaptation representation-engineering activation-steering qwen2-audio
-
Updated
Mar 17, 2026 - Python
Qwen3-0.6B activation steering: style vectors, lens contamination eval, CPRR methodology
python style-transfer mechanistic-interpretability llm-evaluation small-language-models representation-engineering activation-steering qwen3 steering-vectors cprr
-
Updated
Mar 26, 2026 - Python
Evaluation framework of different methods for probing and steering LLMs activations to mitigate Chain-of-Thought Unfaithfulness. Research project by Giovanni M. Occhipinti (University of Bologna), Alessandro Abate e Nandi Schoots (University of Oxford).
chain-of-thought-reasoning representation-engineering activation-probing chain-of-thought-unfaithfulness
-
Updated
Mar 29, 2026 - Python
Phase-aware LLM activation steering and linear probing. A memory-efficient, practical implementation of Representation Engineering (RepE) for safety research.
transformers pytorch ai-safety mechanistic-interpretability llm-safety representation-engineering activation-steering linear-probes
-
Updated
Apr 1, 2026 - Python
LatentBiopsy: Geometric Anomaly Detection for LLM Residual Streams.
-
Updated
Apr 2, 2026 - Python
Representation Rerouting for Agentic Safety: Defending LLM Agents against Prompt Injection via Circuit Breakers and Triplet Loss.
circuit-breakers safety lora triplet-loss adversarial-robustness prompt-injection llm-agents representation-engineering
-
Updated
Apr 9, 2026 - Python
[🏆 CHI26 Best Paper] CoBRA: Reproducible control of LLM agent behavior via classic social science experiments
reproducible-research agent-based-modeling hci representation-learning social-simulation social-science ai-tools ai-agent llm-agent representation-engineering llm-behavior
-
Updated
Apr 21, 2026 - Python
CRSM (Continuous Reasoning State Model): An asynchronous "System 2" architecture that implements Hierarchical State Sovereignty within a Mamba backbone. Unlike traditional search wrappers, CRSM uses Forward-Projected Planning and Sparse-Gated Injection to steer latent manifolds in real-time, decoupling strategic reasoning from token generation.
research reinforcement-learning ai pytorch mcts rl control-theory ssm mamba reasoning state-space-models system-2 large-language-models llm brain-inspired-ai representation-engineering asynchronous-systems
-
Updated
Apr 24, 2026 - Python
Mechanistic interpretability experiments on political control circuits, refusal behavior, concept steering, and late-decoder interactions in open LLMs.
deep-learning transformers alignment censorship interpretability llm mechanistic-interpretability qwen representation-engineering steering-vectors
-
Updated
May 16, 2026 - Python
Coordinate navigation for Transformer internal representations — zero training, any model
transformer pca hidden-states mechanistic-interpretability representation-engineering activation-steering llm-controllability
-
Updated
May 23, 2026 - Python
Functional emotional architecture for LLMs — 42 systems, 1994 tests, 27 psychological theories. Emergent emotions via 7 ANIMA pillars: predictive processing, global workspace, autobiographical memory, ontogenic development, motivational drives, emotional discovery, computational phenomenology.
react python typescript emotion emotions affective-computing fastapi psicology emotional-ai llm big-five representation-engineering steering-vectors appraisal-theory
-
Updated
May 26, 2026 - Python
Lightweight representation engineering dataflow operations for agent developers.
transformers pytorch dataflow parquet huggingface huggingface-transformers duckdb pylance mechanistic-interpretability lancedb transformerlens representation-engineering pragmatic-interpretability
-
Updated
May 27, 2026 - Python
Training-time defense that redistributes LLM refusal via mean/covariance matching + KD, raising linear-ablation attack rank from K=1 to K≥16 (Llama-3.2-1B-Instruct)
pytorch llama ai-alignment adversarial-robustness mechanistic-interpretability llm-safety refusal representation-engineering leace
-
Updated
Jun 12, 2026 - Python
Extending Anthropic's Persona Vectors to a vision-language model: affective image valence covertly shifts an extracted 'evil' direction in LLaVA-1.5-7B, invisibly to text-based harm monitors.
ai-safety interpretability mechanistic-interpretability llava representation-engineering vision-language-models activation-steering persona-vectors
-
Updated
Jun 16, 2026 - Python
Steer2Adapt: data-efficient inference-time LLM adaptation by composing steering vectors via Bayesian optimization over a semantic prior subspace.
bayesian-optimization llm representation-engineering activation-steering llm-adaptation steering-vectors
-
Updated
Jun 30, 2026 - Python
Independent reproduction & mechanistic study of K-Steering (Beyond Linear Steering, arXiv:2505.24535): when/why non-linear classifier-gradient steering beats the CAA linear baseline.
reproduction interpretability llm mechanistic-interpretability representation-engineering activation-steering
-
Updated
Jul 1, 2026 - Python
Local-first activation steering and representation-engineering lab with reproducible ActAdd demos, FastAPI workbench, React UI, and research-grounded limitations.
react pytorch interpretability fastapi huggingface llm llmops representation-engineering activation-steering local-first-ai
-
Updated
Jul 6, 2026 - Python
Add this topic to your repo
To associate your repository with the representation-engineering topic, visit your repo's landing page and select "manage topics."