Elena Tutubalina
Elena Tutubalina on Hugging Face Daily Papers: 15 papers, 9 in the top 3 of their day, 1,134 upvotes.
- Your Transformer Can Hold Two Thoughts at Once: Evidence of Linear Superposition in LLMs 100 upvotes, #2 of 2026-09-25
- The More Popular, The Harder to Forget: Adaptive Popularity for LLM Unlearning 17 upvotes, #9 of 2026-08-20
- Does VLA Even Know the Basics? Measuring Commonsense and World Knowledge Retention in Vision-Language-Action Models 76 upvotes, #3 of 2026-07-01
- Sanity Checks for Sparse Autoencoders: Do SAEs Beat Random Baselines? 55 upvotes, #2 of 2026-02-18
- Global PIQA: Evaluating Physical Commonsense Reasoning Across 100+ Languages and Cultures 9 upvotes, #26 of 2025-10-29
- Emergent Misalignment via In-Context Learning: Narrow in-context examples can produce broadly misaligned LLMs 45 upvotes, #6 of 2025-10-20
- OrtSAE: Orthogonal Sparse Autoencoders Uncover Atomic Features 12 upvotes, #11 of 2025-10-06
- The Rogue Scalpel: Activation Steering Compromises LLM Safety 21 upvotes, #15 of 2025-10-03
- When Punctuation Matters: A Large-Scale Comparison of Prompt Robustness Methods for LLMs 39 upvotes, #7 of 2025-08-19
- Geopolitical biases in LLMs: what are the "good" and the "bad" countries according to contemporary language models 70 upvotes, #1 of 2025-06-11
- I Have Covered All the Bases Here: Interpreting Reasoning Features in Large Language Models via Sparse Autoencoders 110 upvotes, #1 of 2025-03-25
- RuCCoD: Towards Automated ICD Coding in Russian 122 upvotes, #1 of 2025-03-10
- SynthDetoxM: Modern LLMs are Few-Shot Parallel Detoxification Data Annotators 84 upvotes, #2 of 2025-02-11
- CLEAR: Character Unlearning in Textual and Visual Modalities 196 upvotes, #1 of 2024-10-30
- nabla^2DFT: A Universal Quantum Chemistry Dataset of Drug-Like Molecules and a Benchmark for Neural Network Potentials 96 upvotes, #1 of 2024-06-21
Data: hysts-bot-data/daily-papers-stats and the Daily Papers API. Open data: tardellirs/paper-pulse-data. Sister project: Model Pulse, the download history of every model on the Hub.