Profiles2020-2026
Jordan T Ash
Microsoft Research NYC
8Papers
2Last author
13Unique co-authors
2First author
0Solo
First and last author come from byline position, which some venues assign alphabetically.
Recent papers8
- The Coverage Principle: How Pre-Training Enables Post-TrainingAuthor 6 of 8ICLR 2026 · 28 citations
- Self-Improvement in Language Models: The Sharpening MechanismAuthor 6 of 6ICLR 2025
- The Truth is in There: Improving Reasoning in Language Models with Layer-Selective Rank ReductionAuthor 2 of 4ICLR 2024 · 135 citations
- A Study of Plasticity Loss in On-Policy Deep Reinforcement LearningAuthor 2 of 2NeurIPS 2024 · 37 citations
- Exposing Attention Glitches with Flip-Flop Language ModelingAuthor 2 of 5NeurIPS 2023 · 90 citations
- Transformers Learn Shortcuts to AutomataAuthor 2 of 5ICLR 2023 · 11 citations
- Gone Fishing: Neural Active Learning with Fisher EmbeddingsAuthor 1 of 3NeurIPS 2021 · 124 citations
- On Warm-Starting Neural Network TrainingAuthor 1 of 2NeurIPS 2020 · 288 citations
