PrologMCP: A Standardized Prolog Tool Interface for LLM Agents
- lab Hugging Face
- lab arXiv
- location arXivLabs
- model Claude Sonnet 4.6
- model GPT-4.1
- model o4-mini
- person Agnieszka Mensfelt
- product PARARULE-Plus
Researchers have introduced two new systems, PrologMCP and GIST-CMTF, aimed at improving the performance of large language model (LLM) agents in deductive tasks.
PrologMCP is a standardized Prolog tool interface for LLM agents that exposes Prolog as a stateful tool through the Model Context Protocol (MCP)[1]. It features a compact tool interface, structured error reporting, and per-session isolation, making it a reusable primitive for MCP-capable agents. A formalizer agent enhanced with PrologMCP has been shown to match or exceed reasoning LLMs on two subsets of PARARULE-Plus. Meanwhile, GIST-CMTF, a goal-state inference layer, predicts candidate symbolic goals and achieves 97.0% task success, compared with 80.1% for top-goal CMTF and 82.9% for semantic-goal CMTF[2]. GIST-CMTF also reduces wrong-goal execution from 19.4% under top-goal CMTF to 2.5%[2]. Both PrologMCP and GIST-CMTF aim to address the limitations of current LLM agents in handling complex deductive tasks. Frontier reasoning-tuned language models still fail on deductive tasks at depth, according to the PrologMCP research[1].
applicationmodel-releaseinfrastructureresearch-papertool-release
Background sources we checked (8)
- arxiv.org ↗ Frontier reasoning-tuned language models still fail on deductive tasks at depth, and the cost of improved performance through extended internal reasoning scales poorly. Symbolic delegation offers a complementary route: a language model translates the problem, while a solver perfo…
- arxiv.org ↗ We review thirteen generative systems and five supporting datasets for quantum circuit and quantum code generation, identified through a structured scoping review of Hugging Face, arXiv, and provenance tracing (January-February 2026). We organize the field along two axes: artifac…
- huggingface.co ↗ # Paper Pages Paper pages allow people to find artifacts related to a paper such as models, datasets and apps/demos (Spaces). Paper pages also enable the community to discuss about the paper. ## Linking a Paper to a model, dataset or Space If the repository card (`README.md`) …
- huggingface.co ↗ # How to Add a Space to ArXiv ... Demos on Hugging Face Spaces allow a wide audience to try out state-of-the-art machine learning research without writing any code. Hugging Face and ArXiv have collaborated to embed these demos directly along side papers on ArXiv! ... Thanks to th…
- huggingface.co ↗ Daily Papers - Hugging Face new Get trending papers in your email inbox once a day! Get trending papers in your email inbox! Subscribe # Daily Papers ## byAK and the research community - Daily - Weekly - Monthly Trending Papers https://huggingface.co/papers/date/2026-06-…
- en.wikipedia.org ↗ Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co., Ltd., doing business as DeepSeek, is a Chinese artificial intelligence (AI) company that develops large language models (LLMs). Based in Hangzhou, Zhejiang, DeepSeek is owned and funded by High-Flyer, a Chin…
- en.wikipedia.org ↗ A large language model (LLM) is a type of machine learning model designed for natural language processing tasks such as language generation. LLMs are language models with many parameters, and are trained with self-supervised learning on a vast amount of text.…
- en.wikipedia.org ↗ Qwen (also known as Tongyi Qianwen, Chinese: 通义千问; pinyin: Tōngyì Qiānwèn) is a family of large language models developed by Alibaba Cloud. Many Qwen models are distributed under the free and open-source Apache 2.0 license, the source-available Qwen License, or the non-commercial…