{"id":"arxiv:2506.10947","title":"Spurious Rewards: Rethinking Training Signals in RLVR","date":"2025-06-12","section":"Reasoning and the \"physics\" of language models","quality":0.389933,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/","source_url":"https://arxiv.org/abs/2506.10947","text_available":true,"authors":["Rulin Shao","Shuyue Stella Li","Rui Xin"],"outgoing":[{"source":"arxiv:2506.10947","target":"arxiv:2402.03300","evidence":"id","id":"arxiv:2402.03300","title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/636fb871904bc06f742cfdea069796f9e9665f95bd58dc63260eaff5154ef985/"},{"source":"arxiv:2506.10947","target":"arxiv:2407.21783","evidence":"title","id":"arxiv:2407.21783","title":"The Llama 3 Herd of Models","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/de3cebb4c28bc5d41a843279f219dca14476284991ea5c7c0aeab7068a40e306/"},{"source":"arxiv:2506.10947","target":"arxiv:2503.14476","evidence":"title","id":"arxiv:2503.14476","title":"DAPO: An Open-Source LLM Reinforcement Learning System at Scale","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/548e934b912ef84691ace63bf1cfd670e7b0257d9d6986dde926f569e7b2f8b7/"},{"source":"arxiv:2506.10947","target":"arxiv:2503.20783","evidence":"id","id":"arxiv:2503.20783","title":"Understanding R1-Zero-Like Training: A Critical Perspective","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/b49a056f3ce8e269dd65e3b11605927f3c4d0ef074bc68e0b1d7f84e6c9c6c6a/"},{"source":"arxiv:2506.10947","target":"arxiv:2504.13837","evidence":"id","id":"arxiv:2504.13837","title":"Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/b90f70269aeb02311d74ffaadf47e139683424a47118dc18619000ba72174d17/"},{"source":"arxiv:2506.10947","target":"arxiv:2504.16084","evidence":"id","id":"arxiv:2504.16084","title":"TTRL: Test-Time Reinforcement Learning","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/ad54dd58d24afdfefc8aa5087d07c4aab2143e90ef06f6f7db8d3f3333452c48/"},{"source":"arxiv:2506.10947","target":"arxiv:2504.20571","evidence":"id","id":"arxiv:2504.20571","title":"Reinforcement Learning for Reasoning in Large Language Models with One Training Example","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/b2b21c3df49619588c371411dba9467a41f81b3fd4da5441898b5145bb0de120/"},{"source":"arxiv:2506.10947","target":"arxiv:2505.19590","evidence":"id","id":"arxiv:2505.19590","title":"Learning to Reason without External Rewards","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/d8df33b706906a7abf4e660f54c75a81ffe2caaec8b0a10d1a6ad6ec63060832/"},{"source":"arxiv:2506.10947","target":"arxiv:2505.21444","evidence":"id","id":"arxiv:2505.21444","title":"Can Large Reasoning Models Self-Train?","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/206224ca50098e0bdc230140722feec107cfc90c1cef2ed8d3ed52355ef56bca/"},{"source":"arxiv:2506.10947","target":"arxiv:2506.01939","evidence":"id","id":"arxiv:2506.01939","title":"Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoning","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e0f4a55bca612bce5895b6b98f89b0efc14cbd8cf45946a7743347cf6daee1a1/"}],"incoming":[{"source":"arxiv:2504.16084","target":"arxiv:2506.10947","evidence":"id","id":"arxiv:2504.16084","title":"TTRL: Test-Time Reinforcement Learning","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/ad54dd58d24afdfefc8aa5087d07c4aab2143e90ef06f6f7db8d3f3333452c48/"},{"source":"arxiv:2505.19590","target":"arxiv:2506.10947","evidence":"id","id":"arxiv:2505.19590","title":"Learning to Reason without External Rewards","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/d8df33b706906a7abf4e660f54c75a81ffe2caaec8b0a10d1a6ad6ec63060832/"},{"source":"arxiv:2505.21444","target":"arxiv:2506.10947","evidence":"id","id":"arxiv:2505.21444","title":"Can Large Reasoning Models Self-Train?","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/206224ca50098e0bdc230140722feec107cfc90c1cef2ed8d3ed52355ef56bca/"},{"source":"arxiv:2509.25123","target":"arxiv:2506.10947","evidence":"title","id":"arxiv:2509.25123","title":"From $f(x)$ and $g(x)$ to $f(g(x))$: LLMs Learn New Skills in RL by Composing Old Ones","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/7fdfe7abe26ee324fe7abad3ce802148163129a7c157d77bdec859c990868e03/"},{"source":"arxiv:2510.14901","target":"arxiv:2506.10947","evidence":"id","id":"arxiv:2510.14901","title":"Reasoning with Sampling: Your Base Model is Smarter Than You Think","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/46373e6f386387299bbade0a302219a611c3497c5e5b5970e3502f7a5f189687/"},{"source":"arxiv:2512.13961","target":"arxiv:2506.10947","evidence":"id","id":"arxiv:2512.13961","title":"Olmo 3","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/8daa11a9729da9879a7e95c15d4eef3bab8579d2885a14102ed09db111221802/"},{"source":"arxiv:2601.11061","target":"arxiv:2506.10947","evidence":"id","id":"arxiv:2601.11061","title":"Spurious Rewards Paradox: Mechanistically Understanding How RLVR Activates Memorization Shortcuts in LLMs","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/96002027cf4c274573f7320398481a0cd73dbf72f18d7f37c783605f5158ea7b/"}],"text_parts":[{"start":0,"end":15470,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-1.html"},{"start":15470,"end":31070,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-2.html"},{"start":31070,"end":46726,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-3.html"},{"start":46726,"end":62555,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-4.html"},{"start":62555,"end":78531,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-5.html"},{"start":78531,"end":79495,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-6.html"}],"contents":[{"title":"Spurious Rewards: Rethinking Training Signals in RLVR","offset":0,"level":1,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-1.html#text"},{"title":"Abstract","offset":57,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-1.html#text"},{"title":"1 Introduction","offset":1485,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-1.html#text"},{"title":"2 Spurious Rewards Yield Significant RLVR Gains","offset":5615,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-1.html#text"},{"title":"2.1 Experimental Setup","offset":6049,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-1.html#text"},{"title":"2.2 Standard to Weak to Spurious Rewards","offset":7741,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-1.html#text"},{"title":"3 (Lack of) Generalization to Other Models","offset":13179,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-1.html#text"},{"title":"4 Understanding Training Signals from Random Rewards in GRPO","offset":16433,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-2.html#text"},{"title":"5 Case Study: Code Reasoning as a Representative Pre-Existing Behavior","offset":21380,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-2.html#text"},{"title":"5.1 Different Models Exhibit Pre-existing Reasoning Strategy Discrepancies","offset":22612,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-2.html#text"},{"title":"5.2 RLVR with Spurious Rewards Upweight Pre-existing Reasoning Strategies","offset":24106,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-2.html#text"},{"title":"5.3 Intervening Explicitly on Code Reasoning Frequency","offset":25575,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-2.html#text"},{"title":"6 Related Work","offset":27160,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-2.html#text"},{"title":"7 Conclusion","offset":29076,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-2.html#text"},{"title":"Acknowledgments","offset":30059,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-2.html#text"},{"title":"Appendix","offset":30992,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-2.html#text"},{"title":"Appendix A Experimental Setup","offset":31005,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-2.html#text"},{"title":"A.1 Preliminaries on GRPO","offset":31039,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-2.html#text"},{"title":"A.2 Datasets and Models","offset":33262,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-3.html#text"},{"title":"A.3 Training Configuration","offset":34633,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-3.html#text"},{"title":"A.4 Decoding Configuration","offset":35112,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-3.html#text"},{"title":"A.5 Computation Resource","offset":35335,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-3.html#text"},{"title":"A.6 Visualization","offset":35439,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-3.html#text"},{"title":"A.7 Experimental Setups for TTRL and One-Shot RL","offset":35788,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-3.html#text"},{"title":"TTRL setup.","offset":35842,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-3.html#text"},{"title":"Differences in our One-Shot RL setup from ( Wang et al., 2025b ) .","offset":36454,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-3.html#text"},{"title":"Appendix B The Curious Case of Random Rewards","offset":37451,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-3.html#text"},{"title":"Random rewards with varying probabilities consistently improve performance.","offset":38002,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-3.html#text"},{"title":"Recap of notations.","offset":38862,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-3.html#text"},{"title":"B.1 Conjecture: Clipping Bias Brings Training Signals under Random Rewards","offset":40059,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-3.html#text"},{"title":"B.1.1 Expected Advantage of Random Rewards","offset":40139,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-3.html#text"},{"title":"B.1.2 Gradient Derivation of Clipping Bias under Random Rewards","offset":42169,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-3.html#text"},{"title":"Case 1: A ^ ≥ 0 \\hat{A}\\geq 0 .","offset":45411,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-3.html#text"},{"title":"Case 2: A ^ < 0 \\hat{A}<0 .","offset":46444,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-3.html#text"},{"title":"Combining the cases.","offset":47090,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-4.html#text"},{"title":"B.1.3 Clipping Creates Asymmetric Updates Towards Model Prior Knowledge","offset":52558,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-4.html#text"},{"title":"Empirical validation.","offset":54451,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-4.html#text"},{"title":"Why only Qwen-Math models benefit.","offset":56177,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-4.html#text"},{"title":"Implications.","offset":57279,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-4.html#text"},{"title":"B.2 High Stochasticity in RLVR Without Clipping","offset":58042,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-4.html#text"},{"title":"Appendix C Additional Results on AMC","offset":59554,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-4.html#text"},{"title":"Appendix D Additional Results on AIME 2024 and AIME 2025","offset":59802,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-4.html#text"},{"title":"D.1 Spurious Rewards Yield Significant RLVR Gains on Qwen2.5-Math","offset":60551,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-4.html#text"},{"title":"D.2 (Lack of) Generalization to Other Models","offset":61093,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-4.html#text"},{"title":"Appendix E Examination of Existing Methods on Models Beyond Qwen","offset":61857,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-4.html#text"},{"title":"Appendix F Compound Rewards that Inhibit Code Reasoning","offset":62286,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-4.html#text"},{"title":"Inhibiting code reasoning during RLVR with spurious rewards can reduce gains on Qwen2.5-Math-7B, but increase gains on other models.","offset":62415,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-4.html#text"},{"title":"Appendix G Beyond Code Reasoning: Another Beneficial Pattern that RLVR Can Easily Elicit","offset":65113,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-5.html#text"},{"title":"Appendix H Switch of Reasoning Strategies during RLVR","offset":66860,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-5.html#text"},{"title":"Reasoning strategy switches during RLVR.","offset":67051,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-5.html#text"},{"title":"Frequency:","offset":68771,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-5.html#text"},{"title":"Appendix I Spurious Prompts: Qwen2.5-Math-7B’s Unreasonably High Sensitivity to Prompts","offset":70688,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-5.html#text"},{"title":"I.1 Existing Prompts","offset":71371,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-5.html#text"},{"title":"I.2 Spurious Prompts","offset":73083,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-5.html#text"},{"title":"Appendix J Additional Results on Models That Have Undergone RL Training","offset":73924,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-5.html#text"},{"title":"Appendix K Qualitative Analysis on Qwen2.5-Math-7B’s Coding Reasoning Behaviors","offset":75724,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-5.html#text"},{"title":"Qwen2.5-Math-7B can reason in code.","offset":76201,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-5.html#text"},{"title":"Qwen2.5-Math-7B’s code reasoning behavior is robust to numerical perturbations.","offset":76922,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-5.html#text"},{"title":"Qwen2.5-Math-7B’s code reasoning behavior may not directly generalize to rephrased questions.","offset":77838,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-5.html#text"},{"title":"Instructions for reporting errors","offset":78531,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/part-6.html#text"}],"text_code_points":79495,"source_markdown":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f48e13a9656d8eff1882eb80988e07a820c4799401298947deac735c004e8592/source.md"}