{"work":{"id":"3ddd0fd2-c176-408f-9b58-0666c2707f2d","openalex_id":"https://openalex.org/W4405254469","doi":"10.48550/arxiv.2412.06769","arxiv_id":"2412.06769","raw_key":null,"title":"Training Large Language Models to Reason in a Continuous Latent Space","authors":null,"authors_text":"Shibo Hao, Sainbayar Sukhbaatar, DiJia Su, Xian Li, Zhiting Hu, Jason Weston","year":2024,"venue":"cs.CL","abstract":"Large language models (LLMs) are typically constrained to reason in the language space, where they express the reasoning process through a chain-of-thought (CoT) to solve complex problems. However, the language space may not always be optimal for reasoning. Most word tokens primarily ensure textual coherence and are not essential for reasoning, while some critical tokens require complex planning and pose challenges to LLMs. To explore the potential of reasoning beyond language, we introduce a new paradigm called Coconut (Chain of Continuous Thought). Coconut utilizes the last hidden state of the LLM as a representation of the reasoning state, termed \"continuous thought.\" Instead of decoding this state into words, we feed it back to the model as the next input embedding directly in the continuous space. This latent reasoning paradigm enables an advanced reasoning pattern, where continuous thoughts can encode multiple alternative next steps, allowing the model to perform a breadth-first search (BFS) rather than committing prematurely to a single deterministic path as in CoT. Coconut outperforms CoT on logical reasoning tasks that require substantial search during planning and achieves a better trade-off between accuracy and efficiency.","external_url":"https://arxiv.org/abs/2412.06769","cited_by_count":16,"metadata_source":"pith","metadata_fetched_at":"2026-08-05T02:28:24.338817+00:00","pith_arxiv_id":"2412.06769","created_at":"2026-05-09T06:05:36.721375+00:00","updated_at":"2026-08-05T02:28:24.338817+00:00","title_quality_ok":true,"display_title":"Training Large Language Models to Reason in a Continuous Latent Space","render_title":"Training Large Language Models to Reason in a Continuous Latent Space"},"hub":{"state":{"work_id":"3ddd0fd2-c176-408f-9b58-0666c2707f2d","tier":"super_hub","tier_reason":"100+ Pith inbound or 10,000+ external citations","pith_inbound_count":160,"external_cited_by_count":16,"distinct_field_count":14,"first_pith_cited_at":"2024-09-18T15:32:48+00:00","last_pith_cited_at":"2026-07-09T07:54:39+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":"2026-08-20T20:19:24.304269+00:00","tier_text":"super_hub"},"tier":"super_hub","role_counts":[{"context_role":"background","n":32},{"context_role":"method","n":1}],"polarity_counts":[{"context_polarity":"background","n":29},{"context_polarity":"unclear","n":2},{"context_polarity":"support","n":1},{"context_polarity":"use_method","n":1}],"runs":{"ask_index":{"job_type":"ask_index","status":"succeeded","result":{"title":"Training Large Language Models to Reason in a Continuous Latent Space","claims":[{"claim_text":"Large language models (LLMs) are typically constrained to reason in the language space, where they express the reasoning process through a chain-of-thought (CoT) to solve complex problems. However, the language space may not always be optimal for reasoning. Most word tokens primarily ensure textual coherence and are not essential for reasoning, while some critical tokens require complex planning and pose challenges to LLMs. To explore the potential of reasoning beyond language, we introduce a new paradigm called Coconut (Chain of Continuous Thought). Coconut utilizes the last hidden state of t","claim_type":"abstract","evidence_strength":"source_metadata"},{"claim_text":"Self-Verification [107], CoVe [108] Domain MathPrompter [109], PedCoT [110], Physics Reasoner [111] Unstructured Prompt CoT [73], Step-Back, [112] Ask Me Anything [113], CoK [114], SEK [115] Model DeepSeek-R1 [116], Claude 3.7 Sonnet [9], o1 [117] Implicit Quiet-STaR [118], Coconut [119] Planning DEPS [93], ProgPrompt [120], ADaPT [121], ToT [99], RAP [101], TravelPlanner [122], PDDL [123], Mind2Web [124] Figure 2.1: A taxonomy of research on cognition covering different learning and reasoning p","claim_type":"background","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"Although previous studies have shown that distilling knowledge from superior LLMs, regardless of whether supervised fine-tuning (SFT) data are amassed in large quantities or carefully curated [222, 239], can enhance the reasoning abilities of smaller models for Conference acronym 'XX, June 03-05, 2018, Woodstock, NY Trovato et al. solving complex tasks [57, 120, 166], recent studies contend that, merely increasing the volume of SFT data typically yields only a log- linear performance improvement","claim_type":"background","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"ℓ =x ℓ +Attn ℓ (RMSNormℓ (xℓ)) ,(8) xℓ+1 =x ′ ℓ +MLP ℓ RMSNorm′ ℓ (x′ ℓ)\u0001 .(9) RMSNorm itself rescales a hidden vector by its root-mean-square magnitude: RMSNorm(x)=g⊙ x√︃ 1 𝑑 Í𝑑 𝑗=1 𝑥2 𝑗 +𝜖 ,(10) wheregis a learned scale vector. This placement explains why hidden-state norms can accumulate with depth. For a single residual update, write xℓ+1 =x ℓ +u ℓ,u ℓ =𝐹 ℓ (RMSNorm(x ℓ)) ,(11) where𝐹 ℓ denotes the attention or MLP branch. Then ∥xℓ+1 ∥2 2 = ∥xℓ ∥2 2 + ∥uℓ ∥2 2 +2 ⟨xℓ,u ℓ⟩ .(12) Pre-norm cont","claim_type":"background","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"To mitigate this issue, two main approaches have been proposed, primarily differentiated by their intervention requirements during inference. The first category requires no additional intervention during inference, achieving efficiency through prompt engineering [10, 11, 12] or specialized training [13, 14, 15, 16, 17,18] to guide LLMs in generating fewer or even zero [19,20] intermediate tokens during reasoning. The second category operates through real-time token-by-token intervention during i","claim_type":"background","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"Distilling 2-1 [219]; C3oT [78]; TokenSkip [194]; CoT-Valve [130]; Self-Training [133]; Learnto Skip [115]; Token-Budget [58]; Verbosity [72]; Stepwise [31]; Z1 [223]; Prune-on-Logic [243];LS-Mixture SFT [218]; DRP [75]; AutoL2S [125]; Assembly of Experts [79]; Ada-R1 [126];ConCISE [145]; VeriThinker [19]; R1-Compress [187]; CTS [226]; A∗-Thought [205]; TLDR [96];OThink-R1 [235]; PNS [220]; ReCUT [77]; StepEntropy [94]; ASAP [229]; ReasoningOutput-basedEfficient Reasoning LatentRepresentationCom","claim_type":"background","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"extended deliberation depending on task complexity. Developing such cognitively-aligned mechanisms is a key step toward building reasoning agents that are both efficient and reliable. 21 3.6. Perception By bridging visual perception with linguistic abstraction, Large Vision-Language Models (LVLMs) have demonstrated unprecedented capabilities for perceiving and understanding multimodal content [200, 201, 202, 203, 204, 205, 206, 207]. Central to this progress is the incorporation of explicit reas","claim_type":"background","confidence":0.85,"evidence_strength":"citation_context"}],"why_cited":"Pith tracks Training Large Language Models to Reason in a Continuous Latent Space because it crossed a citation-hub threshold. Current citing contexts most often use it as background evidence (30 contexts).","role_counts":[{"n":30,"context_role":"background"},{"n":1,"context_role":"method"}]},"error":null,"updated_at":"2026-06-28T19:27:56.326364+00:00"},"author_expand":{"job_type":"author_expand","status":"succeeded","result":{"authors_linked":[{"id":"2b6e7fd3-2fd2-4b76-9dcc-75699c143546","orcid":null,"display_name":"Shibo Hao"},{"id":"10d99b7a-2002-4541-9ce5-b9f202b2f3f3","orcid":null,"display_name":"Sainbayar Sukhbaatar"},{"id":"1ff92022-6625-4fc6-ae7f-9e770ce863d7","orcid":null,"display_name":"DiJia Su"},{"id":"a36fe16d-621a-4794-bbad-487b1783e650","orcid":null,"display_name":"Xian Li"},{"id":"e35370f1-937f-4584-9db6-f3f0acd0f4c6","orcid":null,"display_name":"Zhiting Hu"},{"id":"a6d6d213-6e2d-4af5-83e4-1b5f5be23d45","orcid":null,"display_name":"Jason Weston"}]},"error":null,"updated_at":"2026-06-28T19:27:56.585638+00:00"},"context_extract":{"job_type":"context_extract","status":"succeeded","result":{"enqueued_papers":25},"error":null,"updated_at":"2026-05-14T16:32:37.356851+00:00"},"graph_features":{"job_type":"graph_features","status":"succeeded","result":{"co_cited":[{"title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","work_id":"e6b75ad5-2877-4168-97c8-710407094d20","shared_citers":19},{"title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","work_id":"c5006563-f3ec-438a-9e35-b7b484f34828","shared_citers":17},{"title":"arXiv preprint arXiv:2405.14838 , year=","work_id":"70541c9d-c3f3-48e4-be00-e7de1622f612","shared_citers":12},{"title":"Compressed chain of thought: Efficient reasoning through dense representations","work_id":"5d72fcbb-d14d-4ac0-8644-50807a64d543","shared_citers":12},{"title":"Training Verifiers to Solve Math Word Problems","work_id":"acab1aa8-b4d6-40e0-a3ee-25341701dca2","shared_citers":12},{"title":"Latent visual reasoning.CoRR, abs/2509.24251","work_id":"b6468cfa-4f13-4e02-b0ec-24ff5cd6785a","shared_citers":11},{"title":"Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach","work_id":"1ee7474f-a930-486e-897c-207b8755f2c9","shared_citers":11},{"title":"Qwen2.5-VL Technical Report","work_id":"69dffacb-bfe8-442d-be86-48624c60426f","shared_citers":10},{"title":"Qwen3 Technical Report","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","shared_citers":10},{"title":"The Llama 3 Herd of Models","work_id":"1549a635-88af-4ac1-acfe-51ae7bb53345","shared_citers":10},{"title":"Think silently, think fast: Dy- namic latent compression of LLM reasoning chains.CoRR, abs/2505.16552","work_id":"125e724a-d7d2-4d72-9bc2-9b4d6dff0b3f","shared_citers":10},{"title":"Qwen2.5 Technical Report","work_id":"d8432992-4980-4a81-85c7-9fa2c2b87f85","shared_citers":9},{"title":"Codi: Compressing chain-of-thought into continuous space via self-distillation","work_id":"c538c1a5-0662-4de7-ac6f-29061c2c9686","shared_citers":8},{"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","shared_citers":8},{"title":"GPT-4o System Card","work_id":"f37bf1c7-4964-4e56-9762-d20da8d9009f","shared_citers":8},{"title":"GPT-4 Technical Report","work_id":"b928e041-6991-4c08-8c81-0359e4097c7b","shared_citers":8},{"title":"Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters","work_id":"a8d50b24-bdf5-46ed-bc4f-2927dfd81f1d","shared_citers":8},{"title":"arXiv preprint arXiv:2404.15758 , year=","work_id":"745f12c5-dbd0-4b89-a2aa-e78d08e61bf1","shared_citers":7},{"title":"Decoupled Weight Decay Regularization","work_id":"07ef7360-d385-4033-83f7-8384a6325204","shared_citers":7},{"title":"InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency","work_id":"b8f5e260-fff5-444e-bcf5-2c42cfefd83d","shared_citers":7},{"title":"LLaVA-OneVision: Easy Visual Task Transfer","work_id":"f5f2452b-f2a9-49ac-b38d-c76e18cdfe49","shared_citers":7},{"title":"Machine mental imagery: Empower multimodal reasoning with latent visual tokens.CoRR, abs/2506.17218","work_id":"d35f1e96-5f12-4e84-990b-e4b05852180e","shared_citers":7},{"title":"Monet: Reasoning in latent visual space beyond images and language","work_id":"ce4ac531-7907-4d8a-afe2-9a0dae21ffa5","shared_citers":7},{"title":"OpenAI o1 System Card","work_id":"68d3c334-0fc9-49e3-b7b0-a69afae933e2","shared_citers":7}],"time_series":[{"n":1,"year":2024},{"n":3,"year":2025},{"n":43,"year":2026}],"dependency_candidates":[]},"error":null,"updated_at":"2026-05-14T16:42:46.372452+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":"2026-05-14T16:32:41.673037+00:00"},"role_polarity":{"job_type":"role_polarity","status":"succeeded","result":{"title":"Training Large Language Models to Reason in a Continuous Latent Space","claims":[{"claim_text":"Large language models (LLMs) are typically constrained to reason in the language space, where they express the reasoning process through a chain-of-thought (CoT) to solve complex problems. However, the language space may not always be optimal for reasoning. Most word tokens primarily ensure textual coherence and are not essential for reasoning, while some critical tokens require complex planning and pose challenges to LLMs. To explore the potential of reasoning beyond language, we introduce a new paradigm called Coconut (Chain of Continuous Thought). Coconut utilizes the last hidden state of t","claim_type":"abstract","evidence_strength":"source_metadata"},{"claim_text":"Self-Verification [107], CoVe [108] Domain MathPrompter [109], PedCoT [110], Physics Reasoner [111] Unstructured Prompt CoT [73], Step-Back, [112] Ask Me Anything [113], CoK [114], SEK [115] Model DeepSeek-R1 [116], Claude 3.7 Sonnet [9], o1 [117] Implicit Quiet-STaR [118], Coconut [119] Planning DEPS [93], ProgPrompt [120], ADaPT [121], ToT [99], RAP [101], TravelPlanner [122], PDDL [123], Mind2Web [124] Figure 2.1: A taxonomy of research on cognition covering different learning and reasoning p","claim_type":"background","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"Although previous studies have shown that distilling knowledge from superior LLMs, regardless of whether supervised fine-tuning (SFT) data are amassed in large quantities or carefully curated [222, 239], can enhance the reasoning abilities of smaller models for Conference acronym 'XX, June 03-05, 2018, Woodstock, NY Trovato et al. solving complex tasks [57, 120, 166], recent studies contend that, merely increasing the volume of SFT data typically yields only a log- linear performance improvement","claim_type":"background","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"ℓ =x ℓ +Attn ℓ (RMSNormℓ (xℓ)) ,(8) xℓ+1 =x ′ ℓ +MLP ℓ RMSNorm′ ℓ (x′ ℓ)\u0001 .(9) RMSNorm itself rescales a hidden vector by its root-mean-square magnitude: RMSNorm(x)=g⊙ x√︃ 1 𝑑 Í𝑑 𝑗=1 𝑥2 𝑗 +𝜖 ,(10) wheregis a learned scale vector. This placement explains why hidden-state norms can accumulate with depth. For a single residual update, write xℓ+1 =x ℓ +u ℓ,u ℓ =𝐹 ℓ (RMSNorm(x ℓ)) ,(11) where𝐹 ℓ denotes the attention or MLP branch. Then ∥xℓ+1 ∥2 2 = ∥xℓ ∥2 2 + ∥uℓ ∥2 2 +2 ⟨xℓ,u ℓ⟩ .(12) Pre-norm cont","claim_type":"background","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"To mitigate this issue, two main approaches have been proposed, primarily differentiated by their intervention requirements during inference. The first category requires no additional intervention during inference, achieving efficiency through prompt engineering [10, 11, 12] or specialized training [13, 14, 15, 16, 17,18] to guide LLMs in generating fewer or even zero [19,20] intermediate tokens during reasoning. The second category operates through real-time token-by-token intervention during i","claim_type":"background","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"Distilling 2-1 [219]; C3oT [78]; TokenSkip [194]; CoT-Valve [130]; Self-Training [133]; Learnto Skip [115]; Token-Budget [58]; Verbosity [72]; Stepwise [31]; Z1 [223]; Prune-on-Logic [243];LS-Mixture SFT [218]; DRP [75]; AutoL2S [125]; Assembly of Experts [79]; Ada-R1 [126];ConCISE [145]; VeriThinker [19]; R1-Compress [187]; CTS [226]; A∗-Thought [205]; TLDR [96];OThink-R1 [235]; PNS [220]; ReCUT [77]; StepEntropy [94]; ASAP [229]; ReasoningOutput-basedEfficient Reasoning LatentRepresentationCom","claim_type":"background","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"extended deliberation depending on task complexity. Developing such cognitively-aligned mechanisms is a key step toward building reasoning agents that are both efficient and reliable. 21 3.6. Perception By bridging visual perception with linguistic abstraction, Large Vision-Language Models (LVLMs) have demonstrated unprecedented capabilities for perceiving and understanding multimodal content [200, 201, 202, 203, 204, 205, 206, 207]. Central to this progress is the incorporation of explicit reas","claim_type":"background","confidence":0.85,"evidence_strength":"citation_context"}],"why_cited":"Pith tracks Training Large Language Models to Reason in a Continuous Latent Space because it crossed a citation-hub threshold. Current citing contexts most often use it as background evidence (30 contexts).","role_counts":[{"n":30,"context_role":"background"},{"n":1,"context_role":"method"}]},"error":null,"updated_at":"2026-06-28T19:27:56.323888+00:00"},"summary_claims":{"job_type":"summary_claims","status":"succeeded","result":{"title":"Training Large Language Models to Reason in a Continuous Latent Space","claims":[{"claim_text":"Large language models (LLMs) are typically constrained to reason in the language space, where they express the reasoning process through a chain-of-thought (CoT) to solve complex problems. However, the language space may not always be optimal for reasoning. Most word tokens primarily ensure textual coherence and are not essential for reasoning, while some critical tokens require complex planning and pose challenges to LLMs. To explore the potential of reasoning beyond language, we introduce a new paradigm called Coconut (Chain of Continuous Thought). Coconut utilizes the last hidden state of t","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks Training Large Language Models to Reason in a Continuous Latent Space because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-14T16:42:48.236978+00:00"}},"summary":{"title":"Training Large Language Models to Reason in a Continuous Latent Space","claims":[{"claim_text":"Large language models (LLMs) are typically constrained to reason in the language space, where they express the reasoning process through a chain-of-thought (CoT) to solve complex problems. However, the language space may not always be optimal for reasoning. Most word tokens primarily ensure textual coherence and are not essential for reasoning, while some critical tokens require complex planning and pose challenges to LLMs. To explore the potential of reasoning beyond language, we introduce a new paradigm called Coconut (Chain of Continuous Thought). Coconut utilizes the last hidden state of t","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks Training Large Language Models to Reason in a Continuous Latent Space because it crossed a citation-hub threshold.","role_counts":[]},"graph":{"co_cited":[{"title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","work_id":"e6b75ad5-2877-4168-97c8-710407094d20","shared_citers":19},{"title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","work_id":"c5006563-f3ec-438a-9e35-b7b484f34828","shared_citers":17},{"title":"arXiv preprint arXiv:2405.14838 , year=","work_id":"70541c9d-c3f3-48e4-be00-e7de1622f612","shared_citers":12},{"title":"Compressed chain of thought: Efficient reasoning through dense representations","work_id":"5d72fcbb-d14d-4ac0-8644-50807a64d543","shared_citers":12},{"title":"Training Verifiers to Solve Math Word Problems","work_id":"acab1aa8-b4d6-40e0-a3ee-25341701dca2","shared_citers":12},{"title":"Latent visual reasoning.CoRR, abs/2509.24251","work_id":"b6468cfa-4f13-4e02-b0ec-24ff5cd6785a","shared_citers":11},{"title":"Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach","work_id":"1ee7474f-a930-486e-897c-207b8755f2c9","shared_citers":11},{"title":"Qwen2.5-VL Technical Report","work_id":"69dffacb-bfe8-442d-be86-48624c60426f","shared_citers":10},{"title":"Qwen3 Technical Report","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","shared_citers":10},{"title":"The Llama 3 Herd of Models","work_id":"1549a635-88af-4ac1-acfe-51ae7bb53345","shared_citers":10},{"title":"Think silently, think fast: Dy- namic latent compression of LLM reasoning chains.CoRR, abs/2505.16552","work_id":"125e724a-d7d2-4d72-9bc2-9b4d6dff0b3f","shared_citers":10},{"title":"Qwen2.5 Technical Report","work_id":"d8432992-4980-4a81-85c7-9fa2c2b87f85","shared_citers":9},{"title":"Codi: Compressing chain-of-thought into continuous space via self-distillation","work_id":"c538c1a5-0662-4de7-ac6f-29061c2c9686","shared_citers":8},{"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","shared_citers":8},{"title":"GPT-4o System Card","work_id":"f37bf1c7-4964-4e56-9762-d20da8d9009f","shared_citers":8},{"title":"GPT-4 Technical Report","work_id":"b928e041-6991-4c08-8c81-0359e4097c7b","shared_citers":8},{"title":"Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters","work_id":"a8d50b24-bdf5-46ed-bc4f-2927dfd81f1d","shared_citers":8},{"title":"arXiv preprint arXiv:2404.15758 , year=","work_id":"745f12c5-dbd0-4b89-a2aa-e78d08e61bf1","shared_citers":7},{"title":"Decoupled Weight Decay Regularization","work_id":"07ef7360-d385-4033-83f7-8384a6325204","shared_citers":7},{"title":"InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency","work_id":"b8f5e260-fff5-444e-bcf5-2c42cfefd83d","shared_citers":7},{"title":"LLaVA-OneVision: Easy Visual Task Transfer","work_id":"f5f2452b-f2a9-49ac-b38d-c76e18cdfe49","shared_citers":7},{"title":"Machine mental imagery: Empower multimodal reasoning with latent visual tokens.CoRR, abs/2506.17218","work_id":"d35f1e96-5f12-4e84-990b-e4b05852180e","shared_citers":7},{"title":"Monet: Reasoning in latent visual space beyond images and language","work_id":"ce4ac531-7907-4d8a-afe2-9a0dae21ffa5","shared_citers":7},{"title":"OpenAI o1 System Card","work_id":"68d3c334-0fc9-49e3-b7b0-a69afae933e2","shared_citers":7}],"time_series":[{"n":1,"year":2024},{"n":3,"year":2025},{"n":43,"year":2026}],"dependency_candidates":[]},"authors":[{"id":"1ff92022-6625-4fc6-ae7f-9e770ce863d7","orcid":null,"display_name":"DiJia Su","source":"manual","import_confidence":0.72},{"id":"a6d6d213-6e2d-4af5-83e4-1b5f5be23d45","orcid":null,"display_name":"Jason Weston","source":"manual","import_confidence":0.72},{"id":"10d99b7a-2002-4541-9ce5-b9f202b2f3f3","orcid":null,"display_name":"Sainbayar Sukhbaatar","source":"manual","import_confidence":0.72},{"id":"2b6e7fd3-2fd2-4b76-9dcc-75699c143546","orcid":null,"display_name":"Shibo Hao","source":"manual","import_confidence":0.72},{"id":"a36fe16d-621a-4794-bbad-487b1783e650","orcid":null,"display_name":"Xian Li","source":"manual","import_confidence":0.72},{"id":"e35370f1-937f-4584-9db6-f3f0acd0f4c6","orcid":null,"display_name":"Zhiting Hu","source":"manual","import_confidence":0.72}]}}