ComprExIT context-compression models and ICAE baselines: 1B/3B/8B, 4x compression, NTP and SFT variants.
Jiangnan Ye
Jiang-nan
AI & ML interests
NLP, Graph Neural Networks, Computer Vision
Recent Activity
upvoted a paper about 11 hours ago
One Assistant, Many Memories: Benchmarking Real-World Group Assistance When Recall Is Not Enough upvoted a paper 10 days ago
On the Opportunities and Challenges of Prefill-to-Autoregressive Cross-Model Transfer upvoted a paper about 1 month ago
Self-Distillation with Verbal RegularizationOrganizations
None yet