[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-how-you-chop-chemistry-papers-changes-what-ai-search-finds":10,"sections":35},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":30,"feedback":34,"feedback_at":22,"cost_usd":34,"total_tokens":34},6945,"how-you-chop-chemistry-papers-changes-what-ai-search-finds","How You Chop Chemistry Papers Changes What AI Search Finds","A new 952 question benchmark on chemistry papers finds embedding model choice, not chunking strategy, drives the biggest retrieval accuracy gains.","A new benchmark built from 952 chemistry question-answer pairs pins down what actually improves AI search over chemistry papers: the embedding model, not how you slice the text.\n\nResearchers built ChemQuests, a corpus of those 952 pairs drawn from 151 ChemRxiv papers spanning 17 chemistry subfields, and used it to construct retrieval benchmarks compatible with the Massive Text Embedding Benchmark. They first screened 41 embedding models on two existing chemistry retrieval benchmarks, ChemNQRetrieval and ChemHotpotQARetrieval, using a geometric-mean metric at rank 10. The strongest performers - retrieval-tuned versions of E5, BGE, and Nomic - were then run through ChemQuests-derived tasks across five chunking strategies, seven chunk sizes, and multiple overlap settings. Across that whole grid, embedding choice produced the largest swings in whether the right evidence got retrieved at all.\n\nFor anyone building retrieval-augmented generation on scientific literature, that is a useful correction to the usual obsession with chunk-size and overlap tuning. The study's practical recommendation - medium-to-large chunks using fixed-token, recursive-token, or hierarchical-section splitting, paired with low overlap - gives teams a concrete starting configuration instead of another parameter sweep to run themselves.\n\nChemistry text is a harsher test than the generic web corpora most embedding models are benchmarked on, thick with symbols, units, and structure-dependent context, so a result that holds up here is a decent signal for other symbol-heavy scientific domains too, not just chemistry.","[\"rag\",\"embeddings\",\"chemistry-ai\",\"benchmarks\"]","2026-09-18T04:00:00.000Z","2026-09-18T23:57:48.994Z","2026-09-18T23:58:00.889Z","published",null,[],"ai",[26,27,28,29],"rag","embeddings","chemistry-ai","benchmarks",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.17277",0,{"sections":36},[37,40,44,49,54,58,62,67,71,76,81,86,91,96],{"name":38,"slug":24,"count":39,"latest_published_at":18},"AI",4082,{"name":41,"slug":42,"count":43,"latest_published_at":18},"Security","security",661,{"name":45,"slug":46,"count":47,"latest_published_at":48},"Policy","policy",339,"2026-09-17T12:00:00.000Z",{"name":50,"slug":51,"count":52,"latest_published_at":53},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":55,"slug":56,"count":57,"latest_published_at":18},"Hardware","hardware",155,{"name":59,"slug":60,"count":61,"latest_published_at":18},"Science","science",125,{"name":63,"slug":64,"count":65,"latest_published_at":66},"Consumer Tech","consumer-tech",99,"2026-09-09T17:27:33.000Z",{"name":68,"slug":69,"count":70,"latest_published_at":18},"Dev Tools","dev-tools",78,{"name":72,"slug":73,"count":74,"latest_published_at":75},"Software","software",75,"2026-09-10T20:41:21.000Z",{"name":77,"slug":78,"count":79,"latest_published_at":80},"Startups","startups",55,"2026-09-09T23:14:29.000Z",{"name":82,"slug":83,"count":84,"latest_published_at":85},"Gaming","gaming",43,"2026-09-10T12:18:06.000Z",{"name":87,"slug":88,"count":89,"latest_published_at":90},"General","general",41,"2026-09-08T01:57:23.000Z",{"name":92,"slug":93,"count":94,"latest_published_at":95},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":97,"slug":98,"count":99,"latest_published_at":100},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]