# Evaluation of Large Language Models via Coupled Token Generation

arXiv: https://arxiv.org/abs/2502.01754 · HTML: https://mlsanity.com/paper/2502.01754/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Builds on 45 of its 63 references that are in the map.
- First posted on arXiv 3 Feb 2025.

## Builds on (most cited first, 30 of 45)

- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Evaluating Large Language Models Trained on Code](https://mlsanity.com/paper/2107.03374/) (2021)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [Finetuned Language Models Are Zero-Shot Learners](https://mlsanity.com/paper/2109.01652/) (2021)
- [Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2204.05862/) (2022)
- [QLoRA: Efficient Finetuning of Quantized LLMs](https://mlsanity.com/paper/2305.14314/) (2023)
- [The Curious Case of Neural Text Degeneration](https://mlsanity.com/paper/1904.09751/) (2019)
- [Self-Instruct: Aligning Language Models with Self-Generated Instructions](https://mlsanity.com/paper/2212.10560/) (2022)
- [Sparks of Artificial General Intelligence: Early experiments with GPT-4](https://mlsanity.com/paper/2303.12712/) (2023)
- [Large Language Models Encode Clinical Knowledge](https://mlsanity.com/paper/2212.13138/) (2022)
- [CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledge](https://mlsanity.com/paper/1811.00937/) (2018)
- [A Survey on Evaluation of Large Language Models](https://mlsanity.com/paper/2307.03109/) (2023)
- [Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference](https://mlsanity.com/paper/2403.04132/) (2024)
- [A General Language Assistant as a Laboratory for Alignment](https://mlsanity.com/paper/2112.00861/) (2021)
- [The Flan Collection: Designing Data and Methods for Effective Instruction Tuning](https://mlsanity.com/paper/2301.13688/) (2023)
- [Cross-Task Generalization via Natural Language Crowdsourcing Instructions](https://mlsanity.com/paper/2104.08773/) (2021)
- [LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methods](https://mlsanity.com/paper/2412.05579/) (2024)
- [LMSYS-Chat-1M: A Large-Scale Real-World LLM Conversation Dataset](https://mlsanity.com/paper/2309.11998/) (2023)
- [PromptSource: An Integrated Development Environment and Repository for Natural Language Prompts](https://mlsanity.com/paper/2202.01279/) (2022)
- [ARES: An Automated Evaluation Framework for Retrieval-Augmented Generation Systems](https://mlsanity.com/paper/2311.09476/) (2023)
- [Generative Judge for Evaluating Alignment](https://mlsanity.com/paper/2310.05470/) (2023)
- [Adding Error Bars to Evals: A Statistical Approach to Language Model Evaluations](https://mlsanity.com/paper/2411.00640/) (2024)
- [Counterfactual Off-Policy Evaluation with Gumbel-Max Structural Causal Models](https://mlsanity.com/paper/1905.05824/) (2019)
- [PRD: Peer Rank and Discussion Improve Large Language Model based Evaluations](https://mlsanity.com/paper/2307.02762/) (2023)
- [LLM-Eval: Unified Multi-Dimensional Automatic Evaluation for Open-Domain Conversations with Large Language Models](https://mlsanity.com/paper/2305.13711/) (2023)
- [Reading Between the Lines: Modeling User Behavior and Costs in AI-Assisted Programming](https://mlsanity.com/paper/2210.14306/) (2022)
- [A Review of the Gumbel-max Trick and its Extensions for Discrete Stochasticity in Machine Learning](https://mlsanity.com/paper/2110.01515/) (2021)

## Built on by (most cited first, 8 of 8)

- [Is Your LLM Overcharging You? Tokenization, Transparency, and Incentives](https://mlsanity.com/paper/2505.21627/) (2025)
- [Compared to What? Baselines and Metrics for Counterfactual Prompting](https://mlsanity.com/paper/2605.01048/) (2026)
- [Large Language Models as Nondeterministic Causal Models](https://mlsanity.com/paper/2509.22297/) (2025)
- [Tokenization Multiplicity Leads to Arbitrary Price Variation in LLM-as-a-service](https://mlsanity.com/paper/2506.06446/) (2025)
- [Precision in Practice: Knowledge Guided Code Summarizing Grounded in Industrial Expectations](https://mlsanity.com/paper/2602.03400/) (2026)
- [Gumbel Machine: Counterfactual Student Writing Generation via Gumbel Noise Steering](https://mlsanity.com/paper/2605.27249/) (2026)
- [Divergence Timing and Cumulative Disagreement under KV-Cache Eviction](https://mlsanity.com/paper/2609.16617/) (2026)
- [agent-risk: Risk Assessment for Changes to AI Agents](https://mlsanity.com/paper/2610.16280/) (2026)

## Abstract (arXiv)

State of the art large language models rely on randomization to respond to a prompt. As an immediate consequence, a model may respond differently to the same prompt if asked multiple times. In this work, we argue that the evaluation and ranking of large language models should control for the randomization underpinning their functioning. Our starting point is the development of a causal model for coupled autoregressive generation, which allows different large language models to sample responses with the same source of randomness. Building upon our causal model, we first show that, on evaluations based on benchmark datasets, coupled autoregressive generation leads to the same conclusions as vanilla autoregressive generation but using provably fewer samples. However, we further show that, on evaluations based on (human) pairwise comparisons, coupled and vanilla autoregressive generation can surprisingly lead to different rankings when comparing more than two models, even with an infinite amount of samples. This suggests that the apparent advantage of a model over others in existing evaluation protocols may not be genuine but rather confounded by the randomness inherent to the generation process. To illustrate and complement our theoretical results, we conduct experiments with several large language models from the Llama, Mistral and Qwen families. We find that, across multiple benchmark datasets, coupled autoregressive generation requires up to 75% fewer samples to reach the same conclusions as vanilla autoregressive generation. Further, we find that the win-rates derived from pairwise comparisons by a strong large language model to prompts from the LMSYS Chatbot Arena platform differ under coupled and vanilla autoregressive generation.