A framework for verifiable reasoning with language models.

framework for detecting hallucinations in LLM chain-of-thought reasoning. Features synthetic data corruption, transformer-based classifiers, Streamlit UI, and FastAPI backend.

ΑΩ+: Speculative reasoning verification for LLMs. Models reasoning as a potential field over token embeddings, directly linking to attention geometry. Implements tetralectic logic & Φ-harmonic stability. Efficient batched implementation with Hutchinson trace estimation, masking for variable-length sequences, and MPS/CPU hybrid for Apple Silicon.