This repository contains the code implementation for the paper "SeSE: Black-Box Uncertainty Quantification for Large Language Models Based on Structural Information Theory", which aims to detect hallucinated content in LLM-generated text.

[ACL 2026 Main] Automated creativity evaluation of LLMs across open-ended tasks via semantic entropy and multi-agent judging.

A native PyTorch GRPO engine for training small reasoning models on consumer GPUs, built to study efficient low-VRAM training and semantic-entropy methods for improving mathematical reasoning.

Structure-first, language-agnostic Translation OS for deterministic, evaluation-ready MT workflows.

Project White Hole — LS7 Natural Operating System (NOS) / 1/7 Framework. Intent Topology using 142857 cyclic parity to reduce semantic entropy in LLMs. Full documentation, 39 proofs, empirical validation, and guided spiral tour.

A per-instance cost model for AI: compute and verification scale with the difficulty of one question.

A lightweight comparative analysis of 3 modern Black-Box Hallucination Detection methods for language models, including SAC3, SelfCheckGPT, and Semantic Entropy.

Desenvolvimento de um sistema multiagentes para auxiliar profissionais fora da área de TI para desenvolver suas próprias soluções tecnológicas, baseadas em seus manuais técnicos com validação de ferramentas de Estatística como Entropia Semântica

HEAT: fact-level hallucination detection for black-box LLMs via semantic entropy and NLI cross-encoders