WTF GENIUS PAPERS Collection Papers that made me appreciate my major and my life a little more. obs=Observation, innov=Innovation. Most papers are abt improving tiny models. • 212 items • Updated about 8 hours ago • 50
Demystifying On-Policy Distillation: Roles, Pathologies, and Regulations Paper • 2607.13399 • Published 4 days ago • 17
Smarter and Cheaper at Once: Byte-Exact KV-Cache Grafting Turns a Frozen Small Model into a Verified-Knowledge Flywheel Paper • 2607.14431 • Published 4 days ago • 10 • 4
Length Penalties Make Chain-of-Thought Less Monitorable Paper • 2607.09786 • Published 11 days ago • 8
WTF GENIUS PAPERS Collection Papers that made me appreciate my major and my life a little more. obs=Observation, innov=Innovation. Most papers are abt improving tiny models. • 212 items • Updated about 8 hours ago • 50
What LLM Forecasters Know but Don't Say: Probing Internal Representations for Calibration and Faithfulness Paper • 2607.08046 • Published 10 days ago • 11
WTF GENIUS PAPERS Collection Papers that made me appreciate my major and my life a little more. obs=Observation, innov=Innovation. Most papers are abt improving tiny models. • 212 items • Updated about 8 hours ago • 50
ShortOPD: Recovering Pruned LLMs with Short-to-Long On-Policy Distillation Paper • 2607.13124 • Published 5 days ago • 17
Towards Autonomous and Auditable Medical Imaging Model Development Paper • 2607.10522 • Published 7 days ago • 17
MuScriptor: An Open Model for Multi-Instrument Music Transcription Paper • 2607.08168 • Published 10 days ago • 19 • 3
MuScriptor: An Open Model for Multi-Instrument Music Transcription Paper • 2607.08168 • Published 10 days ago • 19
Know Before Fix: QA-Driven Repository Knowledge Acquisition for Software Issue Resolution Paper • 2607.11111 • Published 6 days ago • 15
Blind-Spots-Bench: Evaluating Blind Spots in Multimodal Models Paper • 2607.08317 • Published 10 days ago • 30
SynthDocBench: Controlled Benchmark for Long-Context Visual Document Understanding Paper • 2607.10400 • Published 8 days ago • 69
Read It Back: Pretrained MLLMs Are Zero-Shot Reward Models for Text-to-Image Generation Paper • 2607.11886 • Published 6 days ago • 77