OKI TRACE: Local LLM observability. See step-by-step, layer-by-layer what your AI thinks. Logit Lens & Attention for HuggingFace models.

Decoding the black box of LLMs: A comparative analysis of Logit Lens vs. Tuned Lens to interpret intermediate Transformer layers in GPT-2.

Open-source EU AI Act Annex IV compliance toolkit. Mechanistic interpretability + circuit discovery for transformers. One function call generates a court-ready evidence package

🏛️ Champollion cracked hieroglyphs in 1822. I applied the same logic to LLM internals. 95% accuracy, $0 cost, fully reproducible. Contributors welcome.

Find the layer where a language model commits a decision — and steer it. Any open-weight HF model. (WANDERING arc paper #6)

Local Streamlit app for mechanistic interpretability of transformer models.

Mechanistic interpretability CLI for transformer models on Apple Silicon. Analyze per-layer predictions, monitor activation drift, compare models, discover circuits. MLX-based, no GPU needed.

Empirical evidence for predictive coding tendencies in the GPT-2 family: residual stream convergence, activation patching, MLP transform analysis, zero-ablation, and logit lens across 7 languages.

We optimize a compact latent state (frozen weights) to force failed multi-hop chains to output the missing answer D. 5 pre-registered controls show it simply injects D: carries it without the code-fact, leaves intermediates invisible, inert to hop corruption, and doesn’t transfer. No latent composition at 3B (Llama-3.2-3B, Qwen2.5-3B).