Daniil Gavrilov

kefirski's activity

• All
• Models
• Datasets
• Spaces
• Buckets
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

authored 2 papers 5 days ago

Rank-Then-Act: Reward-Free Control from Frame-Order Progress

Paper • 2607.01897 • Published 24 days ago • 7

Qantara: Bridge-Flow Training for Multi-Paradigm JEPA Control

Paper • 2607.04978 • Published 20 days ago • 7

authored 2 papers about 1 month ago

Unstable Features, Reproducible Subspaces: Understanding Seed Dependence in Sparse Autoencoders

Paper • 2606.12138 • Published Jun 10 • 8

Interpreting and Steering a Text-to-Speech Language Model with Sparse Autoencoders

Paper • 2606.10029 • Published Jun 8 • 12

authored 5 papers about 2 months ago

ESSA: Evolutionary Strategies for Scalable Alignment

Paper • 2507.04453 • Published Jul 6, 2025 • 5

Train One Sparse Autoencoder Across Multiple Sparsity Budgets to Preserve Interpretability and Accuracy

Paper • 2505.24473 • Published May 30, 2025

Steering LLM Reasoning Through Bias-Only Adaptation

Paper • 2505.18706 • Published May 24, 2025

Small Vectors, Big Effects: A Mechanistic Study of RL-Induced Reasoning via Steering Vectors

Paper • 2509.06608 • Published Sep 8, 2025

Trust-Region Behavior Blending for On-Policy Distillation

Paper • 2605.31159 • Published May 29 • 69

authored a paper 5 months ago

Next Embedding Prediction Makes World Models Stronger

Paper • 2603.02765 • Published Mar 3 • 21

authored a paper 6 months ago

F-GRPO: Don't Let Your Policy Learn the Obvious and Forget the Rare

Paper • 2602.06717 • Published Feb 6 • 76

authored a paper 12 months ago

Enhancing Vision-Language Model Training with Reinforcement Learning in Synthetic Worlds for Real-World Success

Paper • 2508.04280 • Published Aug 6, 2025 • 35

authored 2 papers about 1 year ago

Teach Old SAEs New Domain Tricks with Boosting

Paper • 2507.12990 • Published Jul 17, 2025 • 12

Train Sparse Autoencoders Efficiently by Utilizing Features Correlation

Paper • 2505.22255 • Published May 28, 2025 • 24

authored 3 papers over 1 year ago

You Do Not Fully Utilize Transformer's Representation Capacity

Paper • 2502.09245 • Published Feb 13, 2025 • 37

Analyze Feature Flow to Enhance Interpretation and Steering in Language Models

Paper • 2502.03032 • Published Feb 5, 2025 • 59

The Differences Between Direct Alignment Algorithms are a Blur

Paper • 2502.01237 • Published Feb 3, 2025 • 113

authored 3 papers almost 2 years ago

Mechanistic Permutability: Match Features Across Layers

Paper • 2410.07656 • Published Oct 10, 2024 • 20

Classifiers are Better Experts for Controllable Text Generation

Paper • 2205.07276 • Published May 15, 2022

PALBERT: Teaching ALBERT to Ponder

Paper • 2204.03276 • Published Apr 7, 2022