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. ā¢ 124 items ā¢ Updated ā¢ 19
Urro PRO
urroxyz
AI & ML interests
computational linguistics major š¤šš
i am autistic. if i come off rude, i probably didn't mean to. please feel free to ask me for clarification.
Recent Activity
updated a collection 1 day ago
WTF GENIUS PAPERS upvoted a paper 1 day ago
RegMix: Data Mixture as Regression for Language Model Pre-training upvoted a paper 1 day ago
SD-E^2: Semantic Exploration for Reasoning Under Token BudgetsOrganizations
updated a
collection 1 day ago
upvoted 14
papers 1 day ago
RegMix: Data Mixture as Regression for Language Model Pre-training
Paper ā¢ 2407.01492 ā¢ Published ā¢ 41
SD-E^2: Semantic Exploration for Reasoning Under Token Budgets
Paper ā¢ 2601.17982 ā¢ Published ā¢ 1
Reasoning Path Divergence: A New Metric and Curation Strategy to Unlock LLM Diverse Thinking
Paper ā¢ 2510.26122 ā¢ Published ā¢ 1
Unlocking Reasoning Capabilities in LLMs via Reinforcement Learning Exploration
Paper ā¢ 2510.03865 ā¢ Published ā¢ 1
Outcome-based Exploration for LLM Reasoning
Paper ā¢ 2509.06941 ā¢ Published ā¢ 1
Controllable Exploration in Hybrid-Policy RLVR for Multi-Modal Reasoning
Paper ā¢ 2602.20197 ā¢ Published ā¢ 1
Wider or Deeper? Scaling LLM Inference-Time Compute with Adaptive Branching Tree Search
Paper ā¢ 2503.04412 ā¢ Published ā¢ 6
Reasoning with Sampling: Your Base Model is Smarter Than You Think
Paper ā¢ 2510.14901 ā¢ Published ā¢ 49
Power-SMC: Low-Latency Sequence-Level Power Sampling for Training-Free LLM Reasoning
Paper ā¢ 2602.10273 ā¢ Published ā¢ 1
Sampling for Quality: Training-Free Reward-Guided LLM Decoding via Sequential Monte Carlo
Paper ā¢ 2604.16453 ā¢ Published ā¢ 1
Diversified Sampling Improves Scaling LLM inference
Paper ā¢ 2502.11027 ā¢ Published ā¢ 1
GuidedSampling: Steering LLMs Towards Diverse Candidate Solutions at Inference-Time
Paper ā¢ 2510.03777 ā¢ Published ā¢ 2
Large Language Monkeys: Scaling Inference Compute with Repeated Sampling
Paper ā¢ 2407.21787 ā¢ Published ā¢ 14
MiMo: Unlocking the Reasoning Potential of Language Model -- From Pretraining to Posttraining
Paper ā¢ 2505.07608 ā¢ Published ā¢ 86
updated a
collection 2 days ago
upvoted a paper 2 days ago
updated a
collection 2 days ago
upvoted a paper 2 days ago
commented on EMO: Pretraining mixture of experts for emergent modularity 2 days ago
I really like this.
It's how MoEs should have always worked, and it's fascinating that it actually succeeds in practice.
Please submit this to Daily Papers so more people can see it:
https://huggingface.co/papers/submit?paperId=2605.06663