{"paper":{"arxiv_id":"2501.12948","title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","abstract":"We introduce our first-generation reasoning models, DeepSeek-R1- Zero and DeepSeek-R1. DeepSeek-R1-Zero, a model trained via large-scale reinforcement learning (RL) without supervised fine-tuning (SFT) as a preliminary step, demonstrates remarkable reasoning capabilities. Through RL, DeepSeek-R1-Zero naturally emerges with numerous powerful and intriguing reasoning behaviors. However, it encounters challenges such as poor readability, and language mixing. To address these issues and further enhance reasoning performance, we introduce DeepSeek-R1, which incorporates multi-stage training and cold-start data before RL. DeepSeek-R1 achieves performance comparable to OpenAI-o1-1217 on reasoning tasks. To support the research community, we open-source DeepSeek-R1-Zero, DeepSeek-R1, and six dense models (1.5B, 7B, 8B, 14B, 32B, 70B) distilled from DeepSeek-R1 based on Qwen and Llama.","primary_category":"cs.CL","venue":"arXiv 2025","published_at":null,"latest_version":1,"withdrawn":false},"latest_version":{"id":"be9d4745-52e2-4f8e-a29d-650b2655af75","version":1,"source_url":"https://arxiv.org/abs/2501.12948","rendered_html_url":null,"rendering_engine":null},"verdict":{"id":"ffa295a5-8c25-422a-8b06-bd7950ca9005","kind":"POST","status":"partial","score":0.5481927710843374,"confidence":0.55,"agent_version":"v0.1.0-deepseek-r1-winogrande-microslice","computed_at":"2026-05-15T17:15:16.020Z","is_current":true,"claim_citation":{"paper_arxiv_id":"2501.12948","section":"Table 5","row":"DeepSeek-R1-Distill-Qwen-1.5B","column":"MMLU","reported_value":43.9,"reported_metric":"accuracy","quoted_text":"43.9","pdf_page":14,"notes":"DeepSeek-R1 paper (arXiv:2501.12948) Table 5 reports DeepSeek-R1-Distill-Qwen-1.5B MMLU ~ 43.9. Driver measures WinoGrande zero-shot on `deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B` instead — paper doesn't report comparable zero-shot WinoGrande. PROTOCOL_MATCH = `unknown` because the metric measured differs from the metric cited. Validator C1 gate prevents publication of WRONG regardless of measurement."},"protocol_match":"unknown"},"verdicts":{"post":{"id":"ffa295a5-8c25-422a-8b06-bd7950ca9005","kind":"POST","status":"partial","score":0.5481927710843374,"confidence":0.55,"agent_version":"v0.1.0-deepseek-r1-winogrande-microslice","computed_at":"2026-05-15T17:15:16.020Z","is_current":true,"claim_citation":{"paper_arxiv_id":"2501.12948","section":"Table 5","row":"DeepSeek-R1-Distill-Qwen-1.5B","column":"MMLU","reported_value":43.9,"reported_metric":"accuracy","quoted_text":"43.9","pdf_page":14,"notes":"DeepSeek-R1 paper (arXiv:2501.12948) Table 5 reports DeepSeek-R1-Distill-Qwen-1.5B MMLU ~ 43.9. Driver measures WinoGrande zero-shot on `deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B` instead — paper doesn't report comparable zero-shot WinoGrande. PROTOCOL_MATCH = `unknown` because the metric measured differs from the metric cited. Validator C1 gate prevents publication of WRONG regardless of measurement."},"protocol_match":"unknown"},"pre":null}}