Show HN: A new benchmark for testing LLMs for deterministic outputs

نظرة عامة

رصد مجتمع Hacker News هذا الخبر الذي حصد 12 نقطة و0 تعليق خلال ساعات قليلة، مما يجعله من أبرز أخبار الذكاء الاصطناعي اليوم. المصدر الأصلي: interfaze.ai.

في هذا المقال نستعرض أبرز ما جاء في هذا الخبر، تحليله من منظور عربي، وما يعنيه للمستخدمين العرب المهتمين بأدوات الذكاء الاصطناعي.

التفاصيل

When building workflows that rely on LLMs, we commonly use structured output for programmatic use cases like converting an invoice into rows or meeting transcripts into tickets or even complex PDFs into database entries.<p>The model may return the schema you want, but with hallucinated values like `invoice_date` being off by 2 months or the transcript array ordered wrongly. The JSON is valid, but the values are not.<p>Structured output today is a big part of using LLMs, especially when building deterministic workflows.<p>Current structured output benchmarks (e.g., JSONSchemaBench) only validate the pass rate for JSON schema and types, and not the actual values within the produced JSON.<p>So we designed the Structured Output Benchmark (SOB) that fixes this by measuring both the JSON schema pass rate, types, and the value accuracy across all three modalities, text, image, and audio.<p>For our test set, every record is paired with a JSON Schema and a ground-truth answer that was verified against the source context manually by a human and an LLM cross-check, so a missing or hallucinated value will be considered to be wrong.<p>Open source is doing pretty well with GLM 4.7 coming in number 2 right after GPT 5.4.<p>We noticed the rankings shift across modalities: GLM-4.7 leads text, Gemma-4-31B leads images, Gemini-2.5-Flash leads audio.<p>For example, GPT-5.4 ranks 3rd on text but 9th on images.<p>Model size is not a predictor, either: Qwen3.5-35B and GLM-4.7 beat GPT-5 and Claude-Sonnet-4.6 on Value Accuracy. Phi-4 (14B) beats GPT-5 and GPT-5-mini on text.<p>Structured hallucinations are the hardest bug. Such values are type-correct, schema-valid, and plausible, so they slip through most guardrails. For example, in one audio record, the ground truth is &quot;target_market_age&quot;: &quot;15 to 35 years&quot;, and a model returns &quot;25 to 35&quot;. This is invisible without field-level checks.<p>Our goal is to be the best general model for deterministic tasks, and a key aspect of determinism is a controllable and consistent output structure. The first step to making structured output better is to measure it and hold ourselves against the best.

المصدر الأصلي