🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

just landed in thailand, presenting referral augmentation ("PageRank x retrieval") at ACL 2024! would love to chat about anything related to retrieval, synthetic data, or code gen :)

ben wrote up extended thoughts on gpt4 attempting usaco with human-in-the-loop! trajectories contain lots of insights, incl. a recurring issue where the model identifies a bug while reflecting but its code revision then fails to fix it 👀

I'm excited to share recent work with Grace Liu and Michael Tang on exploration in RL! A Single Goal is All You Need: Skills and Exploration Emerge from Contrastive RL without Rewards, Demonstrations, or Subgoals Paper, code, and videos: graliuce.github.io/sgcrl/ A thread.

Excited about more robust repository-level code evals!

fast edit 👀

Next-gen evals should be interactive 👀 • Long-horizon env interaction tests realistic planning (SWEBench, WebShop) • Human interaction draws out hidden capabilities (USACO human-in-loop) • Multi-turn interaction with a bad actor better captures the surface area for attacks

"Next frame prediction achieves a PSNR of 29.4, comparable to lossy JPEG compression." 😮

Presenting USACO at Conference on Language Modeling, reach out to chat about anything coding, agents, test-time methods, or other nlp / rl topics :)

👀

If you're at NeurIPS, stop by our workshop oral! Grace Liu will be presenting on single-goal contrastive RL at 2:45-3pm on Sun 12/15, don't miss it :) imol-workshop.github.io/pages/program/

I love this project

at ICLR, come say hi!

VideoGameBench full release! now has metrics 👀