SchemaBridge Fixed Description Constructor - SchemaBench Hallucination: leaderboard

SchemaBridge evaluates relational-schema design, description and reconstruction. Experiment-specific cohorts retain their own populations. Graph edit distance, hallucination, structural F1, text similarity and identifier coverage are distinct measurements. Round-trip runs use a fixed GPT-4o Mini constructor with each named describer. This board reports four-dimensional hallucination ratio (0-1). Lower is better.

Metric: Four-dimensional hallucination ratio (0-1). Source: arxiv.org. Saturation forecast: Estimated already saturated. 5 models tracked. Updated 14 Nov 2026.

Top models

#ModelScoreOverall rank
1Claude Sonnet 4.60.024#165
2GPT-5.4 Mini0.025#398
3CodeLlama-13B-Instruct-hf0.054#9043
4Llama 3.1 8B Instruct0.058#4545
5deepseek-coder-6.7B-instruct0.096#7426

No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.

Interactive version: theaggregate.ai/benchmark?slug=schemabridge-fixed-description-constructor-schemabench-hallucination · How It Works · Data refreshed daily, snapshot 2026-11-16.