🔄 In a Training Loop

Changdae Oh

changdae

·

https://changdaeoh.github.io/

AI & ML interests

Generalization; Distribution Shift; Uncertainty Quantification; Reward Modeling; Post-training

Recent Activity

upvoted a collection 15 days ago

upvoted a paper 17 days ago

LaLa-World: Scaling Latent World Models with Natural Language Actions

upvoted a paper 24 days ago

Into the Unknown: Benchmarking AI agents on open scientific research problems

View all activity

Organizations

upvoted a collection 15 days ago

Muse Glimmer

Muse Glimmer 30B: multimodal agentic model for local deployment. BF16 weights, GGUF k-quants, ExecuTorch builds, DFlash drafter. • 4 items • Updated Aug 10 • 130

upvoted a paper 17 days ago

LaLa-World: Scaling Latent World Models with Natural Language Actions

Paper • 2610.27851 • Published 23 days ago • 41

upvoted a paper 24 days ago

Into the Unknown: Benchmarking AI agents on open scientific research problems

Paper • 2610.19159 • Published about 1 month ago • 56

upvoted a paper 25 days ago

LEASH: Uncertainty-Aware Weak-to-Strong Escalation and Oversight for Computer-Use Agents

Paper • 2610.22280 • Published 29 days ago • 10

upvoted a paper 29 days ago

Harness^2: Recursive Agent Harnessing for an Open World

Paper • 2610.14224 • Published Oct 9 • 69

upvoted 6 papers about 1 month ago

MIMESIS: Learning User Simulators as Training Environments for Interactive Agents

Paper • 2610.09484 • Published Oct 7 • 21

On-Policy Distillation with Negative-Policy Rollouts

Paper • 2610.07874 • Published Oct 6 • 20

DiVeR: Decision-Critical Verifier Learning for VLA Test-Time Scaling

Paper • 2610.04933 • Published Oct 4 • 12

Science or Slop?: Benchmarking and Mitigating Scientific Slop in AI-Generated Papers

Paper • 2610.00531 • Published Sep 30 • 30

Sharpening Tax in Post-Training

Paper • 2610.01509 • Published Oct 1 • 77

AIM: Agentic Idea Management for Automated Research

Paper • 2609.38445 • Published Sep 29 • 32

upvoted a collection 3 months ago

Qwen3.5

21 items • Updated Mar 9 • 1.83k

upvoted a collection 4 months ago

Mistral Large 3

A state-of-the-art, open-weight, general-purpose multimodal model with a granular Mixture-of-Experts architecture. • 4 items • Updated 20 days ago • 106

upvoted 2 papers 4 months ago

Tracing Agentic Failure from the Flow of Success

Paper • 2607.12747 • Published Jul 14 • 17

Multi-Agent LLMs Fail to Explore Each Other

Paper • 2607.11250 • Published Jul 13 • 14

upvoted a collection 4 months ago

Post training

81 items • Updated 6 days ago • 9

upvoted 2 collections 5 months ago

WTF GENIUS PAPERS

Papers that made me appreciate my major and my life a little more. obs=Observation, innov=Innovation. Most papers are abt improving tiny models. • 530 items • Updated 2 days ago • 109

Ministral 3 - Additional Checkpoints

Different formats and Quantized versions of our Ministral 3 family; 14B/8B/3B Instruct/Reasoning GGUF, 3B Instruct ONNX and 14B/8B/3B Instruct BF16. • 10 items • Updated 20 days ago • 34

upvoted 2 papers 5 months ago

Neglected Free Lunch from Post-training: Progress Advantage for LLM Agents

Paper • 2606.26080 • Published Jun 24 • 13

TAPE: Tool-Guided Adaptive Planning and Constrained Execution in Language Model Agents

Paper • 2602.19633 • Published Feb 23 • 10