{"total":47,"papers":[{"id":"arxiv:2508.11408","title":"On-Policy RL Meets Off-Policy Experts: Harmonizing Supervised Fine-Tuning and Reinforcement Learning via Dynamic Weighting","date":"2025-08-15","section":"Data, training, optimization","quality":0.16439,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/48594da2e3f2cdfeb17195dd033d047235fc4b4ffb65030800f64c6faedc4fbe/","source_url":"https://arxiv.org/abs/2508.11408","text_available":true},{"id":"url:paperswithcode.com/paper/optimizing-millions-of-hyperparameters-by","title":"Optimizing Millions of Hyperparameters by Implicit Differentiation","date":"2019-11-06","section":"Data, training, optimization","quality":null,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/c5ceae5fd961be6d0bb1cd78bdf473d8b86954dc0615b3fb35ad9124be5e2643/","source_url":"https://paperswithcode.com/paper/optimizing-millions-of-hyperparameters-by","text_available":false},{"id":"doi:10.1073/pnas.1611835114","title":"Overcoming catastrophic forgetting in neural networks","date":"2017-03-14","section":"Data, training, optimization","quality":0.012852,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/849082368a44d06b93ef3ba95bcfddc6672b7a5e8064ad2b2d19601639266e8b/","source_url":"https://doi.org/10.1073/pnas.1611835114","text_available":true},{"id":"arxiv:2405.20541","title":"Perplexed by Perplexity: Perplexity-Based Data Pruning With Small Reference Models","date":"2024-05-30","section":"Data, training, optimization","quality":0.16386,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/0633300a3d5ba43049372a7b551215c2d1a9443d02b8a74adf639b8a5d8476a8/","source_url":"https://arxiv.org/abs/2405.20541","text_available":true},{"id":"arxiv:2608.27370","title":"Puro-2B: Poor Lab's Qwen2-1.5B Trained on RTX 5090 within $5090","date":"2026-08-27","section":"Data, training, optimization","quality":null,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e7c6bd97e97892c4e1efe65bec9c8dd703a7e817d9544e8b27822583311ab5ab/","source_url":"https://arxiv.org/abs/2608.27370","text_available":true},{"id":"arxiv:2506.08007","title":"Reinforcement Pre-Training","date":"2025-06-09","section":"Data, training, optimization","quality":-0.408418,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/a5ae77d4a2dcd6843e856083ecc6c30a72563e6896503d6efdfae117b8b007e6/","source_url":"https://arxiv.org/abs/2506.08007","text_available":true},{"id":"arxiv:2405.18392","title":"Scaling Laws and Compute-Optimal Training Beyond Fixed Training Durations","date":"2024-05-28","section":"Data, training, optimization","quality":0.24444,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f70e803c548e8a0c524f42ec6203808e4b633b9cc2fb45f4a075af70150f41d1/","source_url":"https://arxiv.org/abs/2405.18392","text_available":true},{"id":"arxiv:2001.08361","title":"Scaling Laws for Neural Language Models","date":"2020-01-23","section":"Data, training, optimization","quality":0.370895,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/82302458a6d03bc051b2e4fdb8277893a81d4bd6a029969045df6a8380904eac/","source_url":"https://arxiv.org/abs/2001.08361","text_available":true},{"id":"arxiv:2210.10760","title":"Scaling Laws for Reward Model Overoptimization","date":"2022-10-19","section":"Data, training, optimization","quality":0.215147,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/fd55acb4c34a2366d8f9c1fc0672a20be0ac4da8643118d0cbf627e891346d2d/","source_url":"https://arxiv.org/abs/2210.10760","text_available":true},{"id":"arxiv:2601.21343","title":"Self-Improving Pretraining: using post-trained models to pretrain better models","date":"2026-01-29","section":"Data, training, optimization","quality":-0.306255,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/43ce1b189d6b21d445c0ce6d4d04b4352d268beb4346da43be19982402befde8/","source_url":"https://arxiv.org/abs/2601.21343","text_available":true},{"id":"arxiv:2010.01412","title":"Sharpness-Aware Minimization for Efficiently Improving Generalization","date":"2020-10-03","section":"Data, training, optimization","quality":0.564643,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e8386ef0dc42519cd17f8346b28a923ab9ad12f8bae3942e216a31c995d97f61/","source_url":"https://arxiv.org/abs/2010.01412","text_available":true},{"id":"arxiv:2305.14342","title":"Sophia: A Scalable Stochastic Second-order Optimizer for Language Model Pre-training","date":"2023-05-23","section":"Data, training, optimization","quality":0.046984,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/afd7fbad6232c6801145c611952e4b6b6aacd1ecb692f56874317ace8a328858/","source_url":"https://arxiv.org/abs/2305.14342","text_available":true},{"id":"arxiv:2401.17401","title":"Step-size Optimization for Continual Learning","date":"2024-01-30","section":"Data, training, optimization","quality":-0.669005,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/c114ace9d4e8c21e4f5e7bf6fdaf29c0df7b9527d2935c401329991ca92e94fd/","source_url":"https://arxiv.org/abs/2401.17401","text_available":true},{"id":"arxiv:1708.07120","title":"Super-Convergence: Very Fast Training of Neural Networks Using Large Learning Rates","date":"2017-08-23","section":"Data, training, optimization","quality":-0.016091,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/604e53f71d7e9128110137811c408ab3b0c041d945f88d001f5095edbfb1a875/","source_url":"https://arxiv.org/abs/1708.07120","text_available":true},{"id":"arxiv:2507.12856","title":"Supervised Fine Tuning on Curated Data is Reinforcement Learning (and can be improved)","date":"2025-07-17","section":"Data, training, optimization","quality":-0.231891,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/5f6c1906eaa67369d9ff15b974756b89141d1b4cf9d7c0f6554e0d56f26e79be/","source_url":"https://arxiv.org/abs/2507.12856","text_available":true},{"id":"arxiv:2302.06675","title":"Symbolic Discovery of Optimization Algorithms","date":"2023-02-13","section":"Data, training, optimization","quality":0.415216,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/3438564652cae2c199823d6d3456a9bf87703e715521d9faaf396c18621f4a5a/","source_url":"https://arxiv.org/abs/2302.06675","text_available":true},{"id":"arxiv:2009.11243","title":"Tasks, stability, architecture, and compute: Training more effective learned optimizers, and using them to train themselves","date":"2020-09-23","section":"Data, training, optimization","quality":0.494615,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/3813d451f50fe60647b2397ec811ddabe99664b1897350808aac27a456914ed2/","source_url":"https://arxiv.org/abs/2009.11243","text_available":true},{"id":"arxiv:2409.03137","title":"The AdEMAMix Optimizer: Better, Faster, Older","date":"2024-09-05","section":"Data, training, optimization","quality":0.319523,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c4e435f36a4ffe28fc019c45f447a6825e58858ce5616b9b4bcd9e9cd245233/","source_url":"https://arxiv.org/abs/2409.03137","text_available":true},{"id":"arxiv:2608.05136","title":"The Loss Does Not See the Basis, but Adam Does","date":"2026-08-05","section":"Data, training, optimization","quality":0.347475,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4b03f7110ecafad60a66ccb0aab5ee9d412451ea522159ec7f58d4b1798b7cfe/","source_url":"https://arxiv.org/abs/2608.05136","text_available":true},{"id":"arxiv:1803.03635","title":"The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks","date":"2018-03-09","section":"Data, training, optimization","quality":0.139358,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/a9d7f0235205afcc709e3e74f28cc7707831bcb1d46239c418cc7e4b624d23bd/","source_url":"https://arxiv.org/abs/1803.03635","text_available":true},{"id":"arxiv:2405.15682","title":"The Road Less Scheduled","date":"2024-05-24","section":"Data, training, optimization","quality":0.34565,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/205e2d6291f9c17e2c0e0de4392fe953a5e711d00b6758c8752a76f69af587ff/","source_url":"https://arxiv.org/abs/2405.15682","text_available":true},{"id":"url:epochai.org/blog/training-compute-of-frontier-ai-models-grows-by-4-5x-per-year","title":"Training compute of frontier models grows 4–5x/year (Epoch AI)","date":"2024-05-28","section":"Data, training, optimization","quality":null,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/71c5b7caa35fdfeb366308b4e7bd2f1a3d8fc0296aae4bee1d69762d38512c65/","source_url":"https://epochai.org/blog/training-compute-of-frontier-ai-models-grows-by-4-5x-per-year","text_available":false},{"id":"doi:10.5281/zenodo.21979851","title":"Why the Third Axis Is Freedom","date":"2026-08-01","section":"Data, training, optimization","quality":null,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/ad1cb25a030788de49889f8170798679149cc66d4b4b53060770d5e4069758e9/","source_url":"https://doi.org/10.5281/zenodo.21979851","text_available":true}],"next":null}