{"work":{"id":"76a0f850-d490-4e4f-ab98-8d25df82cd23","openalex_id":null,"doi":null,"arxiv_id":"2506.13131","raw_key":null,"title":"AlphaEvolve: A coding agent for scientific and algorithmic discovery","authors":null,"authors_text":"Alexander Novikov, Ng\\^an V\\~u, Marvin Eisenberger, Emilien Dupont, Po-Sen Huang, Adam Zsolt Wagner","year":2025,"venue":"cs.AI","abstract":"In this white paper, we present AlphaEvolve, an evolutionary coding agent that substantially enhances capabilities of state-of-the-art LLMs on highly challenging tasks such as tackling open scientific problems or optimizing critical pieces of computational infrastructure. AlphaEvolve orchestrates an autonomous pipeline of LLMs, whose task is to improve an algorithm by making direct changes to the code. Using an evolutionary approach, continuously receiving feedback from one or more evaluators, AlphaEvolve iteratively improves the algorithm, potentially leading to new scientific and practical discoveries. We demonstrate the broad applicability of this approach by applying it to a number of important computational problems. When applied to optimizing critical components of large-scale computational stacks at Google, AlphaEvolve developed a more efficient scheduling algorithm for data centers, found a functionally equivalent simplification in the circuit design of hardware accelerators, and accelerated the training of the LLM underpinning AlphaEvolve itself. Furthermore, AlphaEvolve discovered novel, provably correct algorithms that surpass state-of-the-art solutions on a spectrum of problems in mathematics and computer science, significantly expanding the scope of prior automated discovery methods (Romera-Paredes et al., 2023). Notably, AlphaEvolve developed a search algorithm that found a procedure to multiply two $4 \\times 4$ complex-valued matrices using $48$ scalar multiplications; offering the first improvement, after 56 years, over Strassen's algorithm in this setting. We believe AlphaEvolve and coding agents like it can have a significant impact in improving solutions of problems across many areas of science and computation.","external_url":"https://arxiv.org/abs/2506.13131","cited_by_count":null,"metadata_source":"pith","metadata_fetched_at":"2026-05-25T05:46:39.287541+00:00","pith_arxiv_id":"2506.13131","created_at":"2026-05-09T05:50:26.290415+00:00","updated_at":"2026-05-25T05:46:39.287541+00:00","title_quality_ok":true,"display_title":"AlphaEvolve: A coding agent for scientific and algorithmic discovery","render_title":"AlphaEvolve: A coding agent for scientific and algorithmic discovery"},"hub":{"state":{"work_id":"76a0f850-d490-4e4f-ab98-8d25df82cd23","tier":"super_hub","tier_reason":"100+ Pith inbound or 10,000+ external citations","pith_inbound_count":147,"external_cited_by_count":null,"distinct_field_count":29,"first_pith_cited_at":"2025-07-13T15:21:23+00:00","last_pith_cited_at":"2026-05-22T04:19:04+00:00","author_build_status":"needed","summary_status":"needed","contexts_status":"needed","graph_status":"needed","ask_index_status":"needed","reader_status":"not_needed","recognition_status":"not_needed","updated_at":"2026-06-01T04:13:05.743800+00:00","tier_text":"super_hub"},"tier":"super_hub","role_counts":[{"context_role":"background","n":33},{"context_role":"baseline","n":3},{"context_role":"method","n":3},{"context_role":"dataset","n":2},{"context_role":"other","n":1}],"polarity_counts":[{"context_polarity":"background","n":31},{"context_polarity":"baseline","n":3},{"context_polarity":"use_method","n":3},{"context_polarity":"unclear","n":2},{"context_polarity":"use_dataset","n":2},{"context_polarity":"support","n":1}],"runs":{"ask_index":{"job_type":"ask_index","status":"succeeded","result":{"title":"AlphaEvolve: A coding agent for scientific and algorithmic discovery","claims":[{"claim_text":"In this white paper, we present AlphaEvolve, an evolutionary coding agent that substantially enhances capabilities of state-of-the-art LLMs on highly challenging tasks such as tackling open scientific problems or optimizing critical pieces of computational infrastructure. AlphaEvolve orchestrates an autonomous pipeline of LLMs, whose task is to improve an algorithm by making direct changes to the code. Using an evolutionary approach, continuously receiving feedback from one or more evaluators, AlphaEvolve iteratively improves the algorithm, potentially leading to new scientific and practical d","claim_type":"abstract","evidence_strength":"source_metadata"},{"claim_text":"well and remain robust, but often fall into local optima due to their strict reliance on inaccurate optimizer computational and storage cost estimates. Given these findings, we focus on evolving a de- terministic, code-based policy that executes natively, avoiding the unpredictability of black-box models. We focus on the study's four most competitive heuristic baselines, Extend [59], DB2Advis [72], Anytime [7], and AutoAdmin [9] present the optimal balance be- tween search efficiency and estimat","claim_type":"baseline","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"genetic programming: A survey,\" IEEE Transactions on Evolutionary Computation , vol. 27, no. 3, pp. 621-641, 2022. [7] B. Romera-Paredes, M. Barekatain, A. Novikov, M. Balog, M. P . Kumar, E. Dupont, F. J. Ruiz, J. S. Ellenberg, P . Wang, O. Fawziet al., \"Mathematical discoveries from program search with large language models,\" Nature, vol. 625, no. 7995, pp. 468-475, 2024. [8] A. Novikov, N. V ˜u, M. Eisenberger, E. Dupont, P .-S. Huang, A. Z. Wagner, S. Shirobokov, B. Kozlovskii, F. J. Ruiz, A","claim_type":"background","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"[67] Alexander Novikov, Ngân Vu, Marvin Eisenberger, Emilien Dupont, Po-Sen Huang, Adam Zsolt Wagner, Sergey Shirobokov, Borislav Kozlovskii, Francisco J. R. Ruiz, Ab- bas Mehrabian, M. Pawan Kumar, Abigail See, Swarat Chaudhuri, George Holland, Alex Davies, Sebastian Nowozin, Pushmeet Kohli, and Matej Balog. Alphaevolve: A coding agent for scientific and algorithmic discovery.arXiv preprint arXiv:2506.13131, 2025. [68] OpenAI. Gpt-4 technical report.arXiv preprint arXiv:2303.08774, 2023. [69] O","claim_type":"background","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"OpenEvolve is derived from the comparative result reported in TTT-Discover. The best and second-best results are highlighted in bold and underlined, respectively. Specifically , we found a better construction ( 0.380856) during our ablation study , as indicated by\u0003 (detailed in Section 4.4.1). Method Model Erdős'# Best Human [ 46] - 0.380927 AlphaEvolve [ 89] Gemini-2.0 Pro + Flash 0.380924 AlphaEvolve V2 [ 38] Gemini-2.0 Pro + Flash 0.380924 OpenEvolve [ 6] gpt-oss-120b 0.380965 TTT-Discover [ ","claim_type":"baseline","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"Moving beyond surface-level output metrics, Zhang et al. [36] advocate for internal self-verification by probing hidden states, which empowers models to cease inference upon reaching a predefined confidence threshold [37]. From AutoML to Agentic Discovery.Algorithm discovery has evolved from classical Au- toML [38, 39] to LLM-driven program search, where FunSearch [40], EoH [41], AlphaEvolve [42], and ADAS [43] use LLMs to iteratively propose and refine algorithms in code. Meta-Harness [20] furt","claim_type":"background","confidence":0.8,"evidence_strength":"citation_context"},{"claim_text":"\"Evolution of heuristics: Towards efficient automatic algorithm design using large language model,\"arXiv preprint arXiv:2401.02051, 2024. [39] P. V . T. Dat, L. Doan, and H. T. T. Binh, \"Hsevo: Elevating automatic heuristic design with diversity-driven harmony search and genetic algo- rithm using llms,\" inProceedings of the AAAI Conference on Artificial Intelligence, vol. 39, no. 25, 2025, pp. 26 931-26 938. [40] A. Novikov, N. V ˜u, M. Eisenberger, E. Dupont, P.-S. Huang, A. Z. Wagner, S. Shiro","claim_type":"background","confidence":0.75,"evidence_strength":"citation_context"}],"why_cited":"Pith tracks AlphaEvolve: A coding agent for scientific and algorithmic discovery because it crossed a citation-hub threshold. Current citing contexts most often use it as background evidence (10 contexts).","role_counts":[{"n":10,"context_role":"background"},{"n":2,"context_role":"baseline"}]},"error":null,"updated_at":"2026-05-16T22:39:17.898239+00:00"},"author_expand":{"job_type":"author_expand","status":"succeeded","result":{"authors_linked":[{"id":"2eead753-37ef-4aac-aaa5-5e8be73cc24d","orcid":null,"display_name":"Alexander Novikov"},{"id":"d12e9db7-bf79-47e5-9a97-dfc7db78bcbf","orcid":null,"display_name":"Ng\\^an V\\~u"},{"id":"59958cd2-05a1-439f-a3f5-234178ab26ae","orcid":null,"display_name":"Marvin Eisenberger"},{"id":"8a368ef3-8220-4f10-8613-887950fb10af","orcid":null,"display_name":"Emilien Dupont"},{"id":"3d5d0de8-311c-477d-a959-45ba3a897b9c","orcid":null,"display_name":"Po-Sen Huang"},{"id":"837a3a7d-18ed-4231-9198-08f4602e778d","orcid":null,"display_name":"Adam Zsolt Wagner"}]},"error":null,"updated_at":"2026-05-16T22:39:19.026641+00:00"},"context_extract":{"job_type":"context_extract","status":"succeeded","result":{"enqueued_papers":25},"error":null,"updated_at":"2026-05-14T07:07:23.522075+00:00"},"graph_features":{"job_type":"graph_features","status":"succeeded","result":{"co_cited":[{"title":"Qwen3 Technical Report","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","shared_citers":16},{"title":"GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning","work_id":"40b60d06-dc1c-4799-b75d-ff1eca653049","shared_citers":15},{"title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","work_id":"c5006563-f3ec-438a-9e35-b7b484f34828","shared_citers":12},{"title":"The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery","work_id":"56b6b58d-e73a-4317-896e-36ac5f84e957","shared_citers":12},{"title":"Shinkaevolve: Towards open-ended and sample-efficient program evolution","work_id":"41f30c7f-ee82-49b7-9ce9-463d3124269c","shared_citers":11},{"title":"Thetaevolve: Test-time learning on open problems","work_id":"0c9ee4af-c6aa-47e7-b1cc-863823529ecb","shared_citers":11},{"title":"Learning to discover at test time","work_id":"e6d32375-24bd-4c3a-a7b8-fe20121f31b1","shared_citers":10},{"title":"Codeevolve: An open source evolutionary coding agent for algorithm discovery and optimization","work_id":"b31e9fdb-cb9e-45bb-8e54-e6fc74ca613a","shared_citers":9},{"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","shared_citers":8},{"title":"Pawan Kumar, Emilien Dupont, Francisco J","work_id":"e3e0be4a-249a-44e7-a987-3fb2f9dd3f64","shared_citers":8},{"title":"Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities","work_id":"008df105-2fdd-45d8-857a-8e35868aecb6","shared_citers":7},{"title":"Georgiev, J","work_id":"b3882729-74ba-4d80-9240-bfe03c45c33b","shared_citers":7},{"title":"isolation level","work_id":"c0d964a0-02a1-4ee2-a596-028489ba753f","shared_citers":7},{"title":"SWE-bench: Can Language Models Resolve Real-World GitHub Issues?","work_id":"d0effe15-a689-441a-8e3f-ea35f1c4e4b1","shared_citers":7},{"title":"The AI Scientist-v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Search","work_id":"fa04f346-ee20-4e9d-bf04-3ad3569a8ed1","shared_citers":7},{"title":"Towards an AI co-scientist","work_id":"485486b1-a1a2-4cde-bdda-768930c403e6","shared_citers":7},{"title":"Adae- volve: Adaptive llm driven zeroth-order optimization","work_id":"284cc2bd-522c-4674-99da-c07e54d2afff","shared_citers":6},{"title":"Barbarians at the gate: How ai is upending systems research","work_id":"3b122ca2-f348-411b-89ef-9e7f73a7b00e","shared_citers":6},{"title":"GPT-4 Technical Report","work_id":"b928e041-6991-4c08-8c81-0359e4097c7b","shared_citers":6},{"title":"Meta-Harness: End-to-End Optimization of Model Harnesses","work_id":"5be3c079-4ffa-458f-adcf-204b66f7af51","shared_citers":6},{"title":"Proximal Policy Optimization Algorithms","work_id":"240c67fe-d14d-4520-91c1-38a4e272ca19","shared_citers":6},{"title":"Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models","work_id":"c1d09cf9-9176-403d-8028-630bbc0cbc5d","shared_citers":5},{"title":"Algorithm discovery with llms: Evolutionary search meets reinforcement learning","work_id":"99cb928e-072e-4f23-b18b-e510aa9529fb","shared_citers":5},{"title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","work_id":"e6b75ad5-2877-4168-97c8-710407094d20","shared_citers":5}],"time_series":[{"n":1,"year":2025},{"n":76,"year":2026}],"dependency_candidates":[]},"error":null,"updated_at":"2026-05-14T07:07:27.963213+00:00"},"identity_refresh":{"job_type":"identity_refresh","status":"succeeded","result":{"items":[{"title":"Qwen3 Technical Report","outcome":"unchanged","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","resolver":"local_arxiv","confidence":0.98,"old_work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e"}],"counts":{"fixed":0,"merged":0,"unchanged":1,"quarantined":0,"needs_external_resolution":0},"errors":[],"attempted":1},"error":null,"updated_at":"2026-05-14T07:07:17.300081+00:00"},"role_polarity":{"job_type":"role_polarity","status":"succeeded","result":{"title":"AlphaEvolve: A coding agent for scientific and algorithmic discovery","claims":[{"claim_text":"In this white paper, we present AlphaEvolve, an evolutionary coding agent that substantially enhances capabilities of state-of-the-art LLMs on highly challenging tasks such as tackling open scientific problems or optimizing critical pieces of computational infrastructure. AlphaEvolve orchestrates an autonomous pipeline of LLMs, whose task is to improve an algorithm by making direct changes to the code. Using an evolutionary approach, continuously receiving feedback from one or more evaluators, AlphaEvolve iteratively improves the algorithm, potentially leading to new scientific and practical d","claim_type":"abstract","evidence_strength":"source_metadata"},{"claim_text":"well and remain robust, but often fall into local optima due to their strict reliance on inaccurate optimizer computational and storage cost estimates. Given these findings, we focus on evolving a de- terministic, code-based policy that executes natively, avoiding the unpredictability of black-box models. We focus on the study's four most competitive heuristic baselines, Extend [59], DB2Advis [72], Anytime [7], and AutoAdmin [9] present the optimal balance be- tween search efficiency and estimat","claim_type":"baseline","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"genetic programming: A survey,\" IEEE Transactions on Evolutionary Computation , vol. 27, no. 3, pp. 621-641, 2022. [7] B. Romera-Paredes, M. Barekatain, A. Novikov, M. Balog, M. P . Kumar, E. Dupont, F. J. Ruiz, J. S. Ellenberg, P . Wang, O. Fawziet al., \"Mathematical discoveries from program search with large language models,\" Nature, vol. 625, no. 7995, pp. 468-475, 2024. [8] A. Novikov, N. V ˜u, M. Eisenberger, E. Dupont, P .-S. Huang, A. Z. Wagner, S. Shirobokov, B. Kozlovskii, F. J. Ruiz, A","claim_type":"background","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"[67] Alexander Novikov, Ngân Vu, Marvin Eisenberger, Emilien Dupont, Po-Sen Huang, Adam Zsolt Wagner, Sergey Shirobokov, Borislav Kozlovskii, Francisco J. R. Ruiz, Ab- bas Mehrabian, M. Pawan Kumar, Abigail See, Swarat Chaudhuri, George Holland, Alex Davies, Sebastian Nowozin, Pushmeet Kohli, and Matej Balog. Alphaevolve: A coding agent for scientific and algorithmic discovery.arXiv preprint arXiv:2506.13131, 2025. [68] OpenAI. Gpt-4 technical report.arXiv preprint arXiv:2303.08774, 2023. [69] O","claim_type":"background","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"OpenEvolve is derived from the comparative result reported in TTT-Discover. The best and second-best results are highlighted in bold and underlined, respectively. Specifically , we found a better construction ( 0.380856) during our ablation study , as indicated by\u0003 (detailed in Section 4.4.1). Method Model Erdős'# Best Human [ 46] - 0.380927 AlphaEvolve [ 89] Gemini-2.0 Pro + Flash 0.380924 AlphaEvolve V2 [ 38] Gemini-2.0 Pro + Flash 0.380924 OpenEvolve [ 6] gpt-oss-120b 0.380965 TTT-Discover [ ","claim_type":"baseline","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"Moving beyond surface-level output metrics, Zhang et al. [36] advocate for internal self-verification by probing hidden states, which empowers models to cease inference upon reaching a predefined confidence threshold [37]. From AutoML to Agentic Discovery.Algorithm discovery has evolved from classical Au- toML [38, 39] to LLM-driven program search, where FunSearch [40], EoH [41], AlphaEvolve [42], and ADAS [43] use LLMs to iteratively propose and refine algorithms in code. Meta-Harness [20] furt","claim_type":"background","confidence":0.8,"evidence_strength":"citation_context"},{"claim_text":"\"Evolution of heuristics: Towards efficient automatic algorithm design using large language model,\"arXiv preprint arXiv:2401.02051, 2024. [39] P. V . T. Dat, L. Doan, and H. T. T. Binh, \"Hsevo: Elevating automatic heuristic design with diversity-driven harmony search and genetic algo- rithm using llms,\" inProceedings of the AAAI Conference on Artificial Intelligence, vol. 39, no. 25, 2025, pp. 26 931-26 938. [40] A. Novikov, N. V ˜u, M. Eisenberger, E. Dupont, P.-S. Huang, A. Z. Wagner, S. Shiro","claim_type":"background","confidence":0.75,"evidence_strength":"citation_context"}],"why_cited":"Pith tracks AlphaEvolve: A coding agent for scientific and algorithmic discovery because it crossed a citation-hub threshold. Current citing contexts most often use it as background evidence (10 contexts).","role_counts":[{"n":10,"context_role":"background"},{"n":2,"context_role":"baseline"}]},"error":null,"updated_at":"2026-05-16T22:39:17.902635+00:00"},"summary_claims":{"job_type":"summary_claims","status":"succeeded","result":{"title":"AlphaEvolve: A coding agent for scientific and algorithmic discovery","claims":[{"claim_text":"In this white paper, we present AlphaEvolve, an evolutionary coding agent that substantially enhances capabilities of state-of-the-art LLMs on highly challenging tasks such as tackling open scientific problems or optimizing critical pieces of computational infrastructure. AlphaEvolve orchestrates an autonomous pipeline of LLMs, whose task is to improve an algorithm by making direct changes to the code. Using an evolutionary approach, continuously receiving feedback from one or more evaluators, AlphaEvolve iteratively improves the algorithm, potentially leading to new scientific and practical d","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks AlphaEvolve: A coding agent for scientific and algorithmic discovery because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-14T07:07:27.937775+00:00"}},"summary":{"title":"AlphaEvolve: A coding agent for scientific and algorithmic discovery","claims":[{"claim_text":"In this white paper, we present AlphaEvolve, an evolutionary coding agent that substantially enhances capabilities of state-of-the-art LLMs on highly challenging tasks such as tackling open scientific problems or optimizing critical pieces of computational infrastructure. AlphaEvolve orchestrates an autonomous pipeline of LLMs, whose task is to improve an algorithm by making direct changes to the code. Using an evolutionary approach, continuously receiving feedback from one or more evaluators, AlphaEvolve iteratively improves the algorithm, potentially leading to new scientific and practical d","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks AlphaEvolve: A coding agent for scientific and algorithmic discovery because it crossed a citation-hub threshold.","role_counts":[]},"graph":{"co_cited":[{"title":"Qwen3 Technical Report","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","shared_citers":16},{"title":"GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning","work_id":"40b60d06-dc1c-4799-b75d-ff1eca653049","shared_citers":15},{"title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","work_id":"c5006563-f3ec-438a-9e35-b7b484f34828","shared_citers":12},{"title":"The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery","work_id":"56b6b58d-e73a-4317-896e-36ac5f84e957","shared_citers":12},{"title":"Shinkaevolve: Towards open-ended and sample-efficient program evolution","work_id":"41f30c7f-ee82-49b7-9ce9-463d3124269c","shared_citers":11},{"title":"Thetaevolve: Test-time learning on open problems","work_id":"0c9ee4af-c6aa-47e7-b1cc-863823529ecb","shared_citers":11},{"title":"Learning to discover at test time","work_id":"e6d32375-24bd-4c3a-a7b8-fe20121f31b1","shared_citers":10},{"title":"Codeevolve: An open source evolutionary coding agent for algorithm discovery and optimization","work_id":"b31e9fdb-cb9e-45bb-8e54-e6fc74ca613a","shared_citers":9},{"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","shared_citers":8},{"title":"Pawan Kumar, Emilien Dupont, Francisco J","work_id":"e3e0be4a-249a-44e7-a987-3fb2f9dd3f64","shared_citers":8},{"title":"Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities","work_id":"008df105-2fdd-45d8-857a-8e35868aecb6","shared_citers":7},{"title":"Georgiev, J","work_id":"b3882729-74ba-4d80-9240-bfe03c45c33b","shared_citers":7},{"title":"isolation level","work_id":"c0d964a0-02a1-4ee2-a596-028489ba753f","shared_citers":7},{"title":"SWE-bench: Can Language Models Resolve Real-World GitHub Issues?","work_id":"d0effe15-a689-441a-8e3f-ea35f1c4e4b1","shared_citers":7},{"title":"The AI Scientist-v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Search","work_id":"fa04f346-ee20-4e9d-bf04-3ad3569a8ed1","shared_citers":7},{"title":"Towards an AI co-scientist","work_id":"485486b1-a1a2-4cde-bdda-768930c403e6","shared_citers":7},{"title":"Adae- volve: Adaptive llm driven zeroth-order optimization","work_id":"284cc2bd-522c-4674-99da-c07e54d2afff","shared_citers":6},{"title":"Barbarians at the gate: How ai is upending systems research","work_id":"3b122ca2-f348-411b-89ef-9e7f73a7b00e","shared_citers":6},{"title":"GPT-4 Technical Report","work_id":"b928e041-6991-4c08-8c81-0359e4097c7b","shared_citers":6},{"title":"Meta-Harness: End-to-End Optimization of Model Harnesses","work_id":"5be3c079-4ffa-458f-adcf-204b66f7af51","shared_citers":6},{"title":"Proximal Policy Optimization Algorithms","work_id":"240c67fe-d14d-4520-91c1-38a4e272ca19","shared_citers":6},{"title":"Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models","work_id":"c1d09cf9-9176-403d-8028-630bbc0cbc5d","shared_citers":5},{"title":"Algorithm discovery with llms: Evolutionary search meets reinforcement learning","work_id":"99cb928e-072e-4f23-b18b-e510aa9529fb","shared_citers":5},{"title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","work_id":"e6b75ad5-2877-4168-97c8-710407094d20","shared_citers":5}],"time_series":[{"n":1,"year":2025},{"n":76,"year":2026}],"dependency_candidates":[]},"authors":[{"id":"837a3a7d-18ed-4231-9198-08f4602e778d","orcid":null,"display_name":"Adam Zsolt Wagner","source":"manual","import_confidence":0.72},{"id":"2eead753-37ef-4aac-aaa5-5e8be73cc24d","orcid":null,"display_name":"Alexander Novikov","source":"manual","import_confidence":0.72},{"id":"8a368ef3-8220-4f10-8613-887950fb10af","orcid":null,"display_name":"Emilien Dupont","source":"manual","import_confidence":0.72},{"id":"59958cd2-05a1-439f-a3f5-234178ab26ae","orcid":null,"display_name":"Marvin Eisenberger","source":"manual","import_confidence":0.72},{"id":"d12e9db7-bf79-47e5-9a97-dfc7db78bcbf","orcid":null,"display_name":"Ng\\^an V\\~u","source":"manual","import_confidence":0.72},{"id":"3d5d0de8-311c-477d-a959-45ba3a897b9c","orcid":null,"display_name":"Po-Sen Huang","source":"manual","import_confidence":0.72}]}}