Skip to content
AI & Machine Learning Medium #elo#human-eval#lmsys

Human Preference Elo

Pairwise human votes feed an Elo system. Models rank by who wins more.

A free, animated human preference elo you can read here or embed on any website, from Scrollchart.

Human Preference Elo

Human Preference Elo vs Benchmark ScoreChatbot Arena Elo (pairwise battles) vs MMLU 5-shot accuracy across frontier models112011601200124012801320136065%70%75%80%85%90%95%Elo Rating (human preference)MMLU 5-shot accuracy (%)benchmark-elo gap zonetrendGPT-4o1320Claude 3.51310Gemini 1.5P1270GPT-4 Turbo1248Llama-3 70B1210Mistral Large1195GPT-3.51155Key findingsElo and benchmarkloosely correlatedContamination caninflate BM scoresElo: ~30k+ battlesper leaderboard slotCI bars: 95% bootstrapconfidence intervalsDashed = trend; gapshows human vs statElo requires no benchmark corpus, so contamination is structurally impossible.

A pairwise battle: two model responses to the same prompt, human picks. Wins update Elo ratings. Leaderboard ordering with confidence intervals. Compared with static benchmarks; preference and benchmark scores often diverge.

Good for

  • AI research literacy articles explaining why human preference arenas resist contamination unlike static benchmarks
  • Model selection guides comparing Elo rankings with benchmark accuracy across frontier systems
  • Evaluation methodology content on the limits of MMLU-style benchmarks for measuring real-world usefulness

Source & accuracy

This human preference elo is an editorial illustration built to represent the concept accurately. Where it shows figures, they are typical or representative values chosen to make the relationship clear, not a single underlying dataset. The diagram and its explainer are reviewed and maintained centrally, and updated over time as understanding improves.

Ranking models via pairwise human judgment

Measuring model quality with metrics (BLEU, perplexity) can be misleading. A language model might score well on BLEU but produce stilted, unnatural text. For subjective tasks, human judgment is more reliable. Pairwise evaluation asks humans to compare two model outputs for the same input and pick the better one. Repeat this across many input-output pairs and models.

This pairwise preference data feeds an Elo rating system (borrowed from chess). Each model starts with an equal rating. When a model wins a comparison, its rating increases; when it loses, it decreases. The amount depends on the current ratings: an upset (a lower-rated model beating a higher-rated one) changes ratings more than an expected outcome. After many comparisons, the Elo ratings converge to a stable ranking.

Efficiency and bias in human evaluation

Pairwise evaluation is more efficient than rating each output on an absolute scale (1-5), because humans are better at relative judgment ('A is better than B') than absolute scoring ('A is a 3.7'). However, pairwise evaluation is slower than automated metrics, so sample sizes are smaller.

Bias is a risk. Evaluators might favor longer outputs, outputs from well-known models, or outputs that match their prior beliefs. Anonymization (hiding which model produced which output) and detailed rubrics reduce bias. Disagreement between evaluators is common and often informative: outputs with low inter-rater agreement might be genuinely ambiguous or reflect gaps in the rubric. Modern LLM evaluations increasingly use Elo-based human ratings as the ground truth, especially for open-ended generation tasks.

Embed this diagram

Add this animated human preference elo to your own site. Copy one line of HTML, or use the embed builder for theme and sizing options.

Reference

What this is
A free, embeddable, animated human preference elo for any website.
Who uses it
AI/ML blogs.
How to embed
Copy one line of HTML. No signup. No watermark. Works in WordPress, Webflow, Ghost, Substack, plain HTML.
File size
iframe embed, ~80 KB gzipped (loads on demand, does not block your page paint).
License
Free forever. Editorial explainer text included; updated centrally over time.

Embed format options

Copy the universal HTML snippet, the WordPress shortcode, or an iframe fallback - see the WordPress plugin page for details. Any format keeps the same Core Web Vitals profile and the same explainer text.

Embed snippet
<div data-scrollchart="human-eval-elo" data-scrollchart-v="1"></div>
<script src="https://scrollchart.com/embed.js" async></script>

Frequently asked questions

Where can I get a free animated "Human Preference Elo" for my website?
Scrollchart provides "Human Preference Elo" as a free, embeddable animated diagram you can add to any website with one line of HTML. No signup is required and there is no watermark. The diagram and its explainer text are served from scrollchart.com, so the embed stays current without any maintenance on your end.
How do I embed a human preference elo in a developer or tech blog?
Copy the one-line snippet from the Scrollchart diagram page and paste it into your post HTML. It works in any static site generator, CMS, or hand-coded HTML page. The embed is a thin loader, not an iframe, so the content is fully in your DOM.