COLLIE
COLLIE is a grammar-based framework for systematic construction of constrained text generation tasks. It allows specification of rich, compositional constraints across diverse generation levels and modeling challenges including language understanding, logical reasoning, and semantic planning. The COLLIE-v1 dataset contains 2,080 instances across 13 constraint structures.
GPT-5 from OpenAI currently leads the COLLIE leaderboard with a score of 0.990 across 10 evaluated AI models.
What COLLIE measures
COLLIE is a text benchmark that evaluates large language models on language, reasoning, and writing tasks. LLM Stats tracks 10 models on this benchmark, with a maximum possible score of 1. Current average across reported models is 0.8, with the leader reaching 1.0.
Compare leaders on the best AI for language, best AI for reasoning and best AI for writing leaderboards.
Publication
- Paper
- COLLIE: Systematic Construction of Constrained Text Generation Tasks
- Authors
- Shunyu Yao, Howard Chen, Austin W. Hanjie, Runzhe Yang, and 1 others
- Published
- arXiv
- 2307.08689
Abstract
Text generation under constraints have seen increasing interests in natural language processing, especially with the rapidly improving capabilities of large language models. However, existing benchmarks for constrained generation usually focus on fixed constraint types (e.g.,generate a sentence containing certain words) that have proved to be easy for state-of-the-art models like GPT-4. We present COLLIE, a grammar-based framework that allows the specification of rich, compositional constraints with diverse generation levels (word, sentence, paragraph, passage) and modeling challenges (e.g.,language understanding, logical reasoning, counting, semantic planning). We also develop tools for automatic extraction of task instances given a constraint structure and a raw text corpus. Using COLLIE, we compile the COLLIE-v1 dataset with 2080 instances comprising 13 constraint structures. We perform systematic experiments across five state-of-the-art instruction-tuned language models and analyze their performances to reveal shortcomings. COLLIE is designed to be extensible and lightweight, and we hope the community finds it useful to develop more complex constraints and evaluations in the future.
Progress Over Time
Interactive timeline showing model performance evolution on COLLIE
COLLIE Leaderboard
| Context | Cost | License | ||||
|---|---|---|---|---|---|---|
| 1 |  OpenAI | — | — | — | ||
| 2 | OpenAI | — | — | — | ||
| 3 | OpenAI | — | — | — | ||
| 4 |  Mistral AI | 128B | 256K | $1.50 / $7.50 | ||
| 5 | OpenAI | — | — | — | ||
| 6 | OpenAI | — | 1.0M | $2.00 / $8.00 | ||
| 7 | Mistral AI | 119B | 256K | $0.15 / $0.60 | ||
| 8 | OpenAI | — | 128K | $2.50 / $10.00 | ||
| 9 | OpenAI | — | 1.0M | $0.40 / $1.60 | ||
| 10 | OpenAI | — | 1.0M | $0.10 / $0.40 |
FAQ
Common questions about COLLIE.
More evaluations to explore
Related benchmarks in the same category
A challenging dataset of 448 multiple-choice questions written by domain experts in biology, physics, and chemistry. Questions are Google-proof and extremely difficult, with PhD experts reaching 65% accuracy.
A more robust and challenging multi-task language understanding benchmark that extends MMLU by expanding multiple-choice options from 4 to 10, eliminating trivial questions, and focusing on reasoning-intensive tasks. Features over 12,000 curated questions across 14 domains and causes a 16-33% accuracy drop compared to original MMLU.
All 30 problems from the 2025 American Invitational Mathematics Examination (AIME I and AIME II), testing olympiad-level mathematical reasoning with integer answers from 000-999. Used as an AI benchmark to evaluate large language models' ability to solve complex mathematical problems requiring multi-step logical deductions and structured symbolic reasoning.
Massive Multitask Language Understanding benchmark testing knowledge across 57 diverse subjects including STEM, humanities, social sciences, and professional domains
A verified subset of 500 software engineering problems from real GitHub issues, validated by human annotators for evaluating language models' ability to resolve real-world coding issues by generating patches for Python codebases.
Humanity's Last Exam (HLE) is a multi-modal academic benchmark with 2,500 questions across mathematics, humanities, and natural sciences, designed to test LLM capabilities at the frontier of human knowledge with unambiguous, verifiable solutions