{"total":415,"papers":[{"id":"arxiv:2608.19197","title":"SPADE: Self-Play in Adaptive Synthetic Executable Environments","date":"2026-08-19","section":"Agents, open-endedness, AGI","quality":0.382295,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/c524376d6034b5d79661d6f1a444591404f4a250bcb0790d802c05e02f8249df/","source_url":"https://arxiv.org/abs/2608.19197","text_available":true},{"id":"arxiv:2309.08600","title":"Sparse Autoencoders Find Highly Interpretable Features in Language Models","date":"2023-09-15","section":"AI safety and consciousness","quality":0.096043,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/311cc8e986e42947a9cd36e64cb7dad93d7352e1f5ad495664fce8d0f6fca5da/","source_url":"https://arxiv.org/abs/2309.08600","text_available":true},{"id":"arxiv:2601.11061","title":"Spurious Rewards Paradox: Mechanistically Understanding How RLVR Activates Memorization Shortcuts in LLMs","date":"2026-01-16","section":"Post-training","quality":0.396869,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/96002027cf4c274573f7320398481a0cd73dbf72f18d7f37c783605f5158ea7b/","source_url":"https://arxiv.org/abs/2601.11061","text_available":true},{"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},{"id":"arxiv:2512.02807","title":"SR-GRPO: Stable Rank as an Intrinsic Geometric Reward for Large Language Model Alignment","date":"2025-12-02","section":"Post-training","quality":0.238309,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/ebb3b04baa11b5d0365ca4445da1c2f833108b680ea65272710ed66a57751c0e/","source_url":"https://arxiv.org/abs/2512.02807","text_available":true},{"id":"arxiv:2403.07714","title":"StableToolBench: Towards Stable Large-Scale Benchmarking on Tool Learning of Large Language Models","date":"2024-03-12","section":"Retrieval, embeddings, benchmarks","quality":null,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/ed03ae4fed851402f7c4c5378410fc34374148218bc20e4e7e92522063d06df9/","source_url":"https://arxiv.org/abs/2403.07714","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:2604.19341","title":"Structured Scaling of AI Discovery Across Diverse Scientific Domains","date":"2026-04-21","section":"Harness","quality":0.65138,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/33ad3375247f86197210840a19064548b997e8617475e606d7a7ac2cda64740e/","source_url":"https://arxiv.org/abs/2604.19341","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":"acl:2022.emnlp-main.340","title":"Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+ NLP Tasks","date":"2022-04-16","section":"Retrieval, embeddings, benchmarks","quality":0.101373,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e43a06571b7e8d89abd5f20f8f520c8bc3d3eb32ffb156cb7fc9e0469fc14b0a/","source_url":"https://aclanthology.org/2022.emnlp-main.340/","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:2512.14856","title":"T5Gemma 2: Seeing, Reading, and Understanding Longer","date":"2025-12-16","section":"LLMs: architectures, context, training","quality":-0.289423,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e982edc00d6b6d9e20d11b62004bb8e91131e2d7f717eca944a9914897f273b6/","source_url":"https://arxiv.org/abs/2512.14856","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:2510.13786","title":"The Art of Scaling Reinforcement Learning Compute for LLMs","date":"2025-10-15","section":"Post-training","quality":0.419492,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/72580bcba3d9b32b2f3e83d074f626234b26dc18126edd22a5eacef927cf87da/","source_url":"https://arxiv.org/abs/2510.13786","text_available":true},{"id":"arxiv:2503.02875","title":"The First Few Tokens Are All You Need: An Efficient and Effective Unsupervised Prefix Fine-Tuning Method for Reasoning Models","date":"2025-03-04","section":"Post-training","quality":-0.063911,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/b9349378a2ea40a7998ba364a4587c40dff3cc9b22750e14b3b21e296a4ecb12/","source_url":"https://arxiv.org/abs/2503.02875","text_available":true},{"id":"arxiv:2501.05441","title":"The GAN is dead; long live the GAN! A Modern GAN Baseline","date":"2025-01-09","section":"Self-supervised learning and vision","quality":0.03231,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/a6fcdbec190497e2e52574b7cb2b998385ed1944920c17dc53b710d18f7ba152/","source_url":"https://arxiv.org/abs/2501.05441","text_available":true},{"id":"arxiv:2604.04943","title":"The Illusion of Latent Generalization: Bi-directionality and the Reversal Curse","date":"2026-03-13","section":"Reasoning and the \"physics\" of language models","quality":null,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/14a8648d9d6f2e9700168f1fd2254dd0307f81e5925488becf1a5a72463af956/","source_url":"https://arxiv.org/abs/2604.04943","text_available":true},{"id":"arxiv:2604.11962","title":"The Linear Centroids Hypothesis: Features as Directions Learned by Local Experts","date":"2026-04-13","section":"Representation alignment","quality":null,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/8428fd0cce9753855f4d0f801609b0205a0be52a459d53ae44e7860d00348485/","source_url":"https://arxiv.org/abs/2604.11962","text_available":true},{"id":"arxiv:2407.21783","title":"The Llama 3 Herd of Models","date":"2024-07-31","section":"LLMs: architectures, context, training","quality":0.075302,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/de3cebb4c28bc5d41a843279f219dca14476284991ea5c7c0aeab7068a40e306/","source_url":"https://arxiv.org/abs/2407.21783","text_available":true},{"id":"arxiv:2502.19981","title":"The Lookahead Limitation: Why Multi-Operand Addition is Hard for LLMs","date":"2025-02-27","section":"Reasoning and the \"physics\" of language models","quality":-0.150939,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/fc024e63f0774183f41cf84f17bb5a4238ff43a8ec18a6f8afb9a1abcd113488/","source_url":"https://arxiv.org/abs/2502.19981","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}],"next":"https://vinnik-dmitry07.github.io/favorite-papers/agents/catalog-0-16.html"}