Daniil Gavrilov

35 papers A* 6A 1Journal 25Unranked 3

Year	Rank	Type	Title / Venue / Authors
2026	J	`jnl`	F-GRPO: Don't Let Your Policy Learn the Obvious and Forget the Rare. CoRR Daniil Plyusov, Alexey Gorbatovski, Boris Shaposhnikov, Viacheslav Sinii, Alexey Malakhov, Daniil Gavrilov
2026	J	`jnl`	Next Embedding Prediction Makes World Models Stronger. CoRR George Bredis, Nikita Balagansky, Daniil Gavrilov, Ruslan Rakhimov
2025	A*	`conf`	Analyze Feature Flow to Enhance Interpretation and Steering in Language Models. ICML Daniil Laptev, Nikita Balagansky, Yaroslav Aksenov, Daniil Gavrilov
2025	J	`jnl`	Analyze Feature Flow to Enhance Interpretation and Steering in Language Models. CoRR Daniil Laptev, Nikita Balagansky, Yaroslav Aksenov, Daniil Gavrilov
2025	J	`jnl`	ESSA: Evolutionary Strategies for Scalable Alignment. CoRR Daria Korotyshova, Boris Shaposhnikov, Alexey Malakhov, Nikita Surnachev, George Bredis, Alexey Gorbatovski, Viacheslav Sinii, Daniil Gavrilov
2025	J	`jnl`	Enhancing Vision-Language Model Training with Reinforcement Learning in Synthetic Worlds for Real-World Success. CoRR George Bredis, Stanislav Dereka, Viacheslav Sinii, Ruslan Rakhimov, Daniil Gavrilov
2025	J	`jnl`	Guided Star-Shaped Masked Diffusion. CoRR Viacheslav Meshchaninov, Egor Shibaev, Artem Makoian, Ivan Klimov, Danil Sheshenya, Andrey Malinin, Nikita Balagansky, Daniil Gavrilov, Aibek Alanov, Dmitry P. Vetrov
2025	A*	`conf`	Learn Your Reference Model for Real Good Alignment. ICLR Alexey Gorbatovski, Boris Shaposhnikov, Alexey Malakhov, Nikita Surnachev, Yaroslav Aksenov, Ian Maksimov, Nikita Balagansky, Daniil Gavrilov
2025	A*	`conf`	Mechanistic Permutability: Match Features Across Layers. ICLR Nikita Balagansky, Ian Maksimov, Daniil Gavrilov
2025	J	`jnl`	Small Vectors, Big Effects: A Mechanistic Study of RL-Induced Reasoning via Steering Vectors. CoRR Viacheslav Sinii, Nikita Balagansky, Gleb Gerasimov, Daniil Laptev, Yaroslav Aksenov, Vadim Kurochkin, Alexey Gorbatovski, Boris Shaposhnikov, Daniil Gavrilov
2025	A*	`conf`	Steering LLM Reasoning Through Bias-Only Adaptation. EMNLP Viacheslav Sinii, Alexey Gorbatovski, Artem Cherepanov, Boris Shaposhnikov, Nikita Balagansky, Daniil Gavrilov
2025	J	`jnl`	Steering LLM Reasoning Through Bias-Only Adaptation. CoRR Viacheslav Sinii, Alexey Gorbatovski, Artem Cherepanov, Boris Shaposhnikov, Nikita Balagansky, Daniil Gavrilov
2025	J	`jnl`	Teach Old SAEs New Domain Tricks with Boosting. CoRR Nikita Koriagin, Yaroslav Aksenov, Daniil Laptev, Gleb Gerasimov, Nikita Balagansky, Daniil Gavrilov
2025	J	`jnl`	The Differences Between Direct Alignment Algorithms are a Blur. CoRR Alexey Gorbatovski, Boris Shaposhnikov, Viacheslav Sinii, Alexey Malakhov, Daniil Gavrilov
2025	A*	`conf`	Train One Sparse Autoencoder Across Multiple Sparsity Budgets to Preserve Interpretability and Accuracy. EMNLP Nikita Balagansky, Yaroslav Aksenov, Daniil Laptev, Vadim Kurochkin, Gleb Gerasimov, Nikita Koriagin, Daniil Gavrilov
2025	J	`jnl`	Train One Sparse Autoencoder Across Multiple Sparsity Budgets to Preserve Interpretability and Accuracy. CoRR Nikita Balagansky, Yaroslav Aksenov, Daniil Laptev, Vadim Kurochkin, Gleb Gerasimov, Nikita Koryagin, Daniil Gavrilov
2025	J	`jnl`	Train Sparse Autoencoders Efficiently by Utilizing Features Correlation. CoRR Vadim Kurochkin, Yaroslav Aksenov, Daniil Laptev, Daniil Gavrilov, Nikita Balagansky
2025	J	`jnl`	You Do Not Fully Utilize Transformer's Representation Capacity. CoRR Gleb Gerasimov, Yaroslav Aksenov, Nikita Balagansky, Viacheslav Sinii, Daniil Gavrilov
2024	J	`jnl`	Learn Your Reference Model for Real Good Alignment. CoRR Alexey Gorbatovski, Boris Shaposhnikov, Alexey Malakhov, Nikita Surnachev, Yaroslav Aksenov, Ian Maksimov, Nikita Balagansky, Daniil Gavrilov
2024	—	`conf`	Linear Transformers with Learnable Kernel Functions are Better In-Context Models. ACL (1) Yaroslav Aksenov, Nikita Balagansky, Sofia Maria Lo Cicero Vaina, Boris Shaposhnikov, Alexey Gorbatovski, Daniil Gavrilov
2024	J	`jnl`	Linear Transformers with Learnable Kernel Functions are Better In-Context Models. CoRR Yaroslav Aksenov, Nikita Balagansky, Sofia Maria Lo Cicero Vaina, Boris Shaposhnikov, Alexey Gorbatovski, Daniil Gavrilov
2024	J	`jnl`	Mechanistic Permutability: Match Features Across Layers. CoRR Nikita Balagansky, Ian Maksimov, Daniil Gavrilov
2023	J	`jnl`	Ahead-of-Time P-Tuning. CoRR Daniil Gavrilov, Nikita Balagansky
2023	J	`jnl`	Democratized Diffusion Language Model. CoRR Nikita Balagansky, Daniil Gavrilov
2022	J	`jnl`	Classifiers are Better Experts for Controllable Text Generation. CoRR Askhat Sitdikov, Nikita Balagansky, Daniil Gavrilov, Alexander Markov
2022	J	`jnl`	FastRPB: a Scalable Relative Positional Encoding for Long Sequence Tasks. CoRR Maksim Zubkov, Daniil Gavrilov
2022	J	`jnl`	Linear Interpolation In Parameter Space is Good Enough for Fine-Tuned Language Models. CoRR Mark Rofin, Nikita Balagansky, Daniil Gavrilov
2022	A*	`conf`	PALBERT: Teaching ALBERT to Ponder. NeurIPS Nikita Balagansky, Daniil Gavrilov
2022	—	`conf`	PALBERT: Teaching ALBERT to Ponder. RepL4NLP@ACL Daniil Gavrilov, Nikita Balagansky
2022	J	`jnl`	PALBERT: Teaching ALBERT to Ponder. CoRR Nikita Balagansky, Daniil Gavrilov
2021	A	`conf`	Implicit Unlikelihood Training: Improving Neural Text Generation with Reinforcement Learning. EACL Evgeny Lagutin, Daniil Gavrilov, Pavel Kalaidin
2021	J	`jnl`	Implicit Unlikelihood Training: Improving Neural Text Generation with Reinforcement Learning. CoRR Evgeny Lagutin, Daniil Gavrilov, Pavel Kalaidin
2020	J	`jnl`	Weight Squeezing: Reparameterization for Compression and Fast Inference. CoRR Artem V. Chumachenko, Daniil Gavrilov, Pavel Kalaidin
2019	J	`jnl`	Self-Attentive Model for Headline Generation. CoRR Daniil Gavrilov, Pavel Kalaidin, Valentin Malykh
2019	—	`conf`	Self-attentive Model for Headline Generation. ECIR (2) Daniil Gavrilov, Pavel Kalaidin, Valentin Malykh