Log inSign up
user avatar
Daniel Kang
@ddkang
Distinguished AI research scientist at Bridgewater AIA labs, asst. professor at UIUC CS. Formerly in the Stanford DAWN lab and the Berkeley Sky Lab.
New York, NY
ddkang.github.io
Joined November 2010
102
Following
6,292
Followers
RepliesMedia
  • user avatar
    Daniel Kang
    @ddkang
    Aug 3
    Prior work has shown a counter-intuitive idea that RLVR boosts reasoning using 100% noisy data to a similar level of using clean data. We show that this is false through a more rigorous process of constructing noisy data. We demonstrate that RLVR with noisy data leads to worse
    Image
    Image
    user avatar
    Stella Li
    @StellaLisy
    May 27, 2025
    🤯 We cracked RLVR with... Random Rewards?! Training Qwen2.5-Math-7B with our Spurious Rewards improved MATH-500 by: - Random rewards: +21% - Incorrect rewards: +25% - (FYI) Ground-truth rewards: + 28.8% How could this even work⁉️ Here's why: 🧵 Blogpost: tinyurl.com/spurious-rewar…
  • user avatar
    Daniel Kang
    @ddkang
    Jul 27
    Building text-to-SQL agents? Don’t blindly trust benchmarks/leaderboards. In our VLDB 2026 paper, we find >50% annotation error rates in BIRD and Spider 2.0-Snow, two widely used text-to-SQL benchmarks, shifting agents’ measured performance by up to 19%. 1/9
    Image
    Image
  • user avatar
    Daniel Kang
    @ddkang
    Jul 20
    New research from Bridgewater AIA Labs, UIUC, and MIT: we prove what we believe to be the first non-vacuous generalization bounds for reasoning LLMs on real-world problems. RLVR powers frontier reasoning capabilities yet its generalization to unseen data has remained an open
    Image
  • user avatar
    Daniel Kang
    @ddkang
    Jul 1
    This is a good time to mention that I've joined Bridgewater AIA Labs, with the goal of training the best financial models. I'll also be at ICML 🇰🇷! Reach out if you're interested in chatting, DMs open :)
    user avatar
    Daniel Kang
    @ddkang
    Jun 30
    New research from Bridgewater AIA Labs and @thinkymachines! We show that high quality investor taste is critical for high performance across several investor workflow tasks. Our results point towards a world where organizational expertise leads to models with differentiated