constrained-decoding

Query language for blending SQL and local language models across structured + unstructured data, with type constraints.

Reproduction Package for the paper "Type-Constrained Code Generation with Language Models" [PLDI 2025]

For our ICRA 2025 paper 🏆 "SELP: Generating Safe and Efficient Task Plans for Robot Agents with Large Language Models" by Yi Wu, Zikang Xiong, Yiran Hu, Shreyash Iyengar, Nan Jiang, Aniket Bera, Lin Tan, and Suresh Jagannathan. (🏆 Best Paper Award Finalist!)

Code for paper "Extract, Denoise and Enforce: Evaluating and Improving Concept Preservation for Text-to-Text Generation" EMNLP 2021 and "Constrained Abstractive Summarization: Preserving Factual Consistency with Constrained Generation" arXiv 2020

Constrained Decoding for Diffusion Language Models via Efficient Inference over Finite Automata (NeurIPS 2026)

A tool-calling layer, not a language model — your schemas in, validated calls out, at 48M parameters. Malformed JSON, invented parameter names, and undeclared tools are structurally unreachable on any catalog. Adapt it to your own catalog; 11 negative results included.

[Pytorch] Efficient tokenization for recommendations and generative retrieval. Inspired by STATIC decoding from "Vectorizing the Trie"

Speculative grammar backtracking algorithm for LLM decoding conforming to some lark context-free grammar (CFG)

Every language before it was built for humans to read and learn. LOVA is the first built for AI: programs are integer sequences — 64 one-byte operators, structured faults with repair hints, declared effects, budgets, lineage, a Rust VM at 11M steps/s, an MCP server. Open source: extend, modify or rebuild it to make AI faster.

Beyond token-by-token agents. JevSpawn enables parallel action exploration through adaptive action spaces, without additional training.

Context-Free Grammar-guided Generation of FHIR Resources Using Large Language Models (MIE 2026)

VibeDrift - Run any LLM on your own hardware. Bypass the VRAM wall with CPU/RAM inference, MOE expert offloading, and 4-bit quantization. No Cloud, no Subscription.

🤖 AI-powered function calling engine using a LLM to convert natural language prompts into validated JSON function calls.

A function calling tool that translates natural language prompts into structured function calls using a small LLM and constrained decoding

Artifact-backed evaluation of how structured-output contracts change LLM semantics, validity, and execution outcomes.

Lexical + graph-augmented RAG over vLLM, using grammar-constrained decoding (Outlines) to extract a Neo4j knowledge graph with a 0.6B model.

Grammar-aided Constrained Decondig for self-aligned LLM with SFT and GRPO

Constrained decoding system for generating valid function calls with a language model. Developed at 42 Málaga.

Deterministic Procedural JSON Generation via a 0.6B Parameter Language Model

Reproduction & re-implementation of the AloLab paper (arXiv:2605.02363): closing the structured-output reliability gap in small LLMs via iterative black-box prompt optimization. GSM8K, 4 models, ablations, McNemar significance.