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

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

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.

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

Refuse the unsafe step, cap the runaway cost. A Nozick-grounded gate for LLM agents (Truth-Tracking + Token Budget Contract).

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