FormBench：配方专利细粒度检索基准
原标题：FormBench: A Benchmark for Fine-Grained Retrieval in Formulation Patents
AI 摘要
研究者发布 FormBench，一个面向配方专利细粒度检索的基准，包含 55,352 条查询和两个各约 100 万段落的语料变体（随机干扰项 C0 与结构相近干扰项 C1），并提供基于材料本体的分级相关性标注。他们评测了 OpenAI、Voyage 等商业 API 编码器、Nomic、GTE-large、Qwen3-Embed 等开源编码器及其任务自适应微调版本，以及 BM25 基线。结果显示商业 API 在 C1 上 nDCG@10 仅 0.009，远低于 BM25 的 0.375，而 TAPT 微调的 Qwen3-Embed 达到 0.4085；作者将 API 失败归因于域内几何坍缩。
正文节选
FormBench: A Benchmark for Fine-Grained Retrieval in Formulation Patents Abstract We introduce FormBench, a platform for testing the ability of models to retrieve precise chemical information from technical documents. FormBench contains 55,352 queries and two retrieval corpus variants each with 1M passages: FormBench-Random (C0), with randomly sampled distractors, and FormBench-Structured (C1), with structurally proximate distractors. Each query is paired with an anchor passage, a hard negative,