DailyPapers

@HuggingPapers

Tweeting interesting papers submitted at huggingface.co/papers. Submit your own at hf.co/papers/submit, and link models/datasets/demos to it!

Anywhere

DailyPapers
@HuggingPapers
29m
ScienceBuddy: recursive-in-recursive self-improvement for interactive scientific agents An interactive scientific workspace from Gen-Verse where researcher feedback becomes tasks and rubrics, letting agents improve both their harness and the model behind it.
1
DailyPapers
@HuggingPapers
4h
StepAudio 3 Music StepFun's new long-form music generation model plans structure with ABC-CoT, then renders up to 5m30s of 48-kHz stereo with a flow-matching DiT.
1
DailyPapers
@HuggingPapers
8h
StepAudio 3 Realtime Technical Report StepFun introduces a full-duplex audio-language model that reasons while speaking, achieving top scores on MMSU (90.6) and AA Full-Duplex Bench (98.9).
2
DailyPapers
@HuggingPapers
12h
AI for Games in the Foundation Model Era A 120-page survey from NUS and NTU mapping six roles for AI across the game lifecycle: play and act, model players and games, design, build and maintain, generate and adapt at runtime, test and evaluate.
1
DailyPapers
@HuggingPapers
16h
Recursive self-improvement beyond novelty A new survey introduces the Headroom-Closed Index (HCI) and an L1-L5 roadmap for how AI systems can genuinely improve themselves, from executing predefined routines to revising their own improvement mechanisms.
2