Open-source library for analyzing entropy and uncertainty in LLM outputs. Dual entropy analysis, reasoning chain inspection, visual attention patterns.

python · transformers · interpretability

Hallucination detection for RAG systems using a fine-tuned BERT classifier.

python · bert · rag · hallucination detection

MCP server for reading and debugging MCP logs directly inside the client.

typescript · mcp · debugging

Full 3D muon scattering tomography of reinforced concrete using Geant4 simulation and neural-network-based event reduction. x85 event reduction.

jinst · 2025

2D reconstruction for cultural heritage with x45 event reduction and x9 resolution improvement using deep learning on synthetic muography data.

epia 2025 · springer

2D reconstruction of 1cm iron bars inside 30cm concrete blocks via muography + deep learning. Dramatic improvement in image quality and acquisition time.

jinst · arxiv 2502.03339 · 2025

GPT4-based MultiParameters benchmark for code LLMs. Outperforms GPT-3.5 and WizardCoder on code readability.

arxiv 2309.11385 · sep 2023

Open-source C# code generation model built on CodeLlama 34B using EvolInstruct. 56.33% on MultiPL-E benchmark.

arxiv 2311.03243 · nov 2023