[EMNLP '23] Discriminator-Guided Chain-of-Thought Reasoning

一本人生编程的教科书——当老天是 Python 程序员

[ICML 2023] Answering Complex Logical Queries on Knowledge Graphs via Query Computation Tree Optimization

A symbolic benchmark for verifiable chain-of-thought financial reasoning. Includes executable templates, 58 topics across 12 domains, and ChainEval metrics.

Synthetic data generation for evaluating LLM symbolic and logic reasoning

Code and data for Paper "Enhancing Ethical Explanations of Large Language Models through Iterative Symbolic Refinement"

[EMNLP 2024] NormTab: Improving Symbolic Reasoning in LLMs Through Tabular Data Normalization

Complex-valued neuro-symbolic transformer using PyTorch.

Symbolic Reasoning Task environment for training RL agents. Inspired OpenAI Gymnasium and Formal Methods

Contradiction-native research runtime for verifier-backed ontology revision, structural abstraction invention, and replayable proof artifacts.

Non-Western symbolic reasoning benchmark for LLMs. Multi-step rule-following inference within the Ba Zi (八字) formal system. Frozen lookup tables, Python reference implementation, mechanically verified gold CoT cases.

PyTorch to Bend converter - transforms neural networks into parallel functional code using Interaction Nets for HVM2

Interaction Net Equivalence Testing for LLMs

Pure-Python baseline solver for Kaggle ARC Prize 2025 on ARC-AGI-2

Symbolic Transformers: 2.2MB models for logical reasoning. Achieves 47% accuracy with 566K parameters—220× smaller than GPT-2. Proves data quality > model size for symbolic AI. 🔬 Novel base-625 symbolic encoding | 🚀 Edge-deployable | 📊 Open research

Prototype reasoning engine for rule-based decision-making in AIOp

🧠 Linguaggio Cognitivo Ibrido per Agenti AI Trasparenti | Hybrid AI Language combining symbolic reasoning with LLMs

Low Resource Language Reasoning Engine - Enterprise Grade Symbolic NLP. 100% accuracy across 5 languages, 16ms latency, NO LLMs, pure symbolic reasoning.

Verifier-backed abstraction invention for small formal protocol and concurrency systems with contradiction-driven ontology revision

Pure program synthesis solver for ARC-AGI-2 — 74/1000=7.4%, no neural networks, no LLMs