{"id":"https://openalex.org/W7220901878","doi":"https://doi.org/10.64898/2026.10.01.755314","title":"Evaluating Autonomous LLM Agents Across Molecular Prediction and Optimization Benchmarks","display_name":"Evaluating Autonomous LLM Agents Across Molecular Prediction and Optimization Benchmarks","publication_year":2026,"publication_date":"2026-10-08","ids":{"openalex":"https://openalex.org/W7220901878","doi":"https://doi.org/10.64898/2026.10.01.755314"},"language":"en","primary_location":{"id":"doi:10.64898/2026.10.01.755314","doi":"https://doi.org/10.64898/2026.10.01.755314","is_oa":true,"landing_page_url":"https://doi.org/10.64898/2026.10.01.755314","pdf_url":"https://www.biorxiv.org/content/biorxiv/early/2026/10/07/2026.10.01.755314.full.pdf","source":{"id":"https://openalex.org/S4306402567","display_name":"bioRxiv (Cold Spring Harbor Laboratory)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I2750212522","host_organization_name":"Cold Spring Harbor Laboratory","host_organization_lineage":["https://openalex.org/I2750212522"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"acceptedVersion","is_accepted":true,"is_published":false,"raw_source_name":"bioRxiv","raw_type":"posted-content"},"volume":null,"type":"preprint","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://www.biorxiv.org/content/biorxiv/early/2026/10/07/2026.10.01.755314.full.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5114403521","display_name":"Khachik Smbatyan","orcid":"https://orcid.org/0000-0002-1661-0044","observed_orcids":["https://orcid.org/0000-0002-1661-0044"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Khachik Smbatyan","raw_affiliation_strings":["Deep Origin Inc., South San Francisco, CA 94080, USA"],"raw_orcid":"https://orcid.org/0000-0002-1661-0044","affiliations":[{"raw_affiliation_string":"Deep Origin Inc., South San Francisco, CA 94080, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5067485758","display_name":"Tsolak Ghukasyan","orcid":"https://orcid.org/0000-0003-2389-517X","observed_orcids":["https://orcid.org/0000-0003-2389-517X"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tsolak Ghukasyan","raw_affiliation_strings":["Deep Origin Inc., South San Francisco, CA 94080, USA"],"raw_orcid":"https://orcid.org/0000-0003-2389-517X","affiliations":[{"raw_affiliation_string":"Deep Origin Inc., South San Francisco, CA 94080, USA","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5114403522","display_name":"Garik Petrosyan","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Garik Petrosyan","raw_affiliation_strings":["Deep Origin Inc., South San Francisco, CA 94080, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Deep Origin Inc., South San Francisco, CA 94080, USA","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":["https://openalex.org/A5114403522"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10211","display_name":"Computational Drug Discovery Methods","score":0.9976000189781189,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10211","display_name":"Computational Drug Discovery Methods","score":0.9976000189781189,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11948","display_name":"Machine Learning in Materials Science","score":0.0010999999940395355,"subfield":{"id":"https://openalex.org/subfields/2505","display_name":"Materials Chemistry"},"field":{"id":"https://openalex.org/fields/25","display_name":"Materials Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T13702","display_name":"Machine Learning in Healthcare","score":9.999999747378752e-05,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/llm-based-agents","display_name":"LLM agents","score":0.875},{"id":"https://openalex.org/keywords/molecular-property-prediction","display_name":"molecular property prediction","score":0.8289999961853027},{"id":"https://openalex.org/keywords/admet-prediction","display_name":"ADMET prediction","score":0.8090000152587891},{"id":"https://openalex.org/keywords/molecular-optimization","display_name":"molecular optimization","score":0.7789999842643738},{"id":"https://openalex.org/keywords/codex","display_name":"codices","score":0.6690000295639038},{"id":"https://openalex.org/keywords/drug-discovery","display_name":"drug discovery","score":0.5979999899864197},{"id":"https://openalex.org/keywords/benchmark-evaluation","display_name":"benchmark evaluation","score":0.4480000138282776}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6442000269889832},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.5127999782562256},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4871000051498413},{"id":"https://openalex.org/C60478076","wikidata":"https://www.wikidata.org/wiki/Q3036835","display_name":"Reference data","level":2,"score":0.46650001406669617},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.45089998841285706},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4465000033378601},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.43709999322891235},{"id":"https://openalex.org/C188154048","wikidata":"https://www.wikidata.org/wiki/Q6803609","display_name":"Mean absolute error","level":3,"score":0.35589998960494995},{"id":"https://openalex.org/C167085575","wikidata":"https://www.wikidata.org/wiki/Q6803654","display_name":"Mean squared prediction error","level":2,"score":0.3093999922275543},{"id":"https://openalex.org/C2987595161","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Optimization algorithm","level":2,"score":0.2565999925136566},{"id":"https://openalex.org/C45804977","wikidata":"https://www.wikidata.org/wiki/Q7239673","display_name":"Predictive modelling","level":2,"score":0.2554999887943268}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.64898/2026.10.01.755314","doi":"https://doi.org/10.64898/2026.10.01.755314","is_oa":true,"landing_page_url":"https://doi.org/10.64898/2026.10.01.755314","pdf_url":"https://www.biorxiv.org/content/biorxiv/early/2026/10/07/2026.10.01.755314.full.pdf","source":{"id":"https://openalex.org/S4306402567","display_name":"bioRxiv (Cold Spring Harbor Laboratory)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I2750212522","host_organization_name":"Cold Spring Harbor Laboratory","host_organization_lineage":["https://openalex.org/I2750212522"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"acceptedVersion","is_accepted":true,"is_published":false,"raw_source_name":"bioRxiv","raw_type":"posted-content"}],"best_oa_location":{"id":"doi:10.64898/2026.10.01.755314","doi":"https://doi.org/10.64898/2026.10.01.755314","is_oa":true,"landing_page_url":"https://doi.org/10.64898/2026.10.01.755314","pdf_url":"https://www.biorxiv.org/content/biorxiv/early/2026/10/07/2026.10.01.755314.full.pdf","source":{"id":"https://openalex.org/S4306402567","display_name":"bioRxiv (Cold Spring Harbor Laboratory)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I2750212522","host_organization_name":"Cold Spring Harbor Laboratory","host_organization_lineage":["https://openalex.org/I2750212522"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"acceptedVersion","is_accepted":true,"is_published":false,"raw_source_name":"bioRxiv","raw_type":"posted-content"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7220901878.pdf","grobid_xml":"https://content.openalex.org/works/W7220901878.grobid-xml"},"referenced_works_count":20,"referenced_works":["https://openalex.org/W1536680647","https://openalex.org/W2909063104","https://openalex.org/W3093934881","https://openalex.org/W4399940463","https://openalex.org/W4399990919","https://openalex.org/W4402467035","https://openalex.org/W4404986804","https://openalex.org/W4407385137","https://openalex.org/W4415242186","https://openalex.org/W4416234289","https://openalex.org/W4416920970","https://openalex.org/W4417524783","https://openalex.org/W4417536917","https://openalex.org/W7091542513","https://openalex.org/W7164414229","https://openalex.org/W7165625761","https://openalex.org/W7172426212","https://openalex.org/W7172443136","https://openalex.org/W7172446914","https://openalex.org/W7172450735"],"related_works":["https://openalex.org/W4416034548","https://openalex.org/W7077063820","https://openalex.org/W4405185373","https://openalex.org/W7162148170","https://openalex.org/W7130831421","https://openalex.org/W7162689699","https://openalex.org/W4406030356","https://openalex.org/W7204181022","https://openalex.org/W7164345841","https://openalex.org/W7161167326"],"abstract_inverted_index":{"Abstract":[0],"Large":[1],"language":[2],"model":[3],"(LLM)":[4],"agents":[5,215],"are":[6],"increasingly":[7],"capable":[8],"of":[9,55,96,152],"carrying":[10],"out":[11],"autonomous":[12,35],"computational":[13],"research,":[14],"but":[15],"it":[16],"remains":[17],"unclear":[18],"whether":[19],"they":[20],"can":[21,216],"develop":[22,218],"molecular":[23,177,200,220],"modeling":[24],"methods":[25,182,190,224],"that":[26,179,212],"compete":[27],"with":[28,143],"strong":[29],"human-developed":[30,237],"approaches.":[31,238],"Here,":[32],"we":[33],"evaluate":[34],"method":[36],"development":[37],"across":[38,87,225],"four":[39],"settings:":[40],"Therapeutics":[41],"Data":[42],"Commons":[43],"(TDC)":[44],"ADMET":[45],"tasks,":[46],"the":[47,51,56,62,70,83,91,101,115,139,144,166,185],"OpenADMET":[48,57],"ExpansionRx":[49],"Challenge,":[50,60],"activity":[52],"prediction":[53,221],"track":[54],"PXR":[58],"Induction":[59],"and":[61,75,99,104,124,134,222],"Practical":[63],"Molecular":[64],"Optimization":[65],"(PMO)":[66],"benchmark.":[67,80],"Agents":[68],"followed":[69],"prescribed":[71],"data":[72,202],"splits,":[73],"metrics,":[74],"evaluation":[76],"protocols":[77],"for":[78],"each":[79],"On":[81,121,155,171],"TDC,":[82],"best":[84,167],"agent-developed":[85],"score":[86],"configurations":[88,112,127],"improved":[89,113],"on":[90,94,100,107,114,118,150],"leaderboard":[92,103],"reference":[93,106,117],"18":[95],"21":[97],"tasks":[98],"combined":[102,116],"peer-reviewed":[105],"17,":[108],"while":[109],"individual":[110],"Codex":[111,126,158,174],"13":[119],"tasks.":[120],"ExpansionRx,":[122],"five-agent":[123,145,157,173],"single-agent":[125],"achieved":[128,159],"overall":[129],"scores":[130,194],"corresponding":[131],"to":[132,138],"fourth":[133],"sixth":[135],"place":[136],"relative":[137],"published":[140,168],"final":[141],"leaderboard,":[142],"configuration":[146],"reaching":[147],"top-five":[148],"performance":[149],"seven":[151],"nine":[153],"endpoints.":[154],"PXR,":[156],"a":[160,176,204],"mean":[161],"absolute":[162],"error":[163],"slightly":[164],"outperforming":[165],"challenge":[169],"entry.":[170],"PMO,":[172],"developed":[175],"optimizer":[178],"outperformed":[180],"existing":[181],"evaluated":[183],"under":[184,195],"benchmark\u2019s":[186],"standard":[187],"setup;":[188],"other":[189],"obtained":[191],"higher":[192],"reported":[193],"modified":[196],"conditions,":[197],"including":[198],"additional":[199],"pretraining":[201],"or":[203,234],"different":[205,227],"oracle-call":[206],"budget.":[207],"Together,":[208],"these":[209],"results":[210],"show":[211],"current":[213],"LLM":[214],"autonomously":[217],"high-performing":[219],"optimization":[223],"substantially":[226],"drug-discovery":[228],"settings,":[229],"in":[230],"several":[231],"cases":[232],"matching":[233],"exceeding":[235],"leading":[236]},"counts_by_year":[],"updated_date":"2026-11-04T10:54:05.112534","created_date":"2026-10-09T00:00:00"}