Log inSign up
user avatar
Joykirat
@joykiratsingh
CS PhD Student @unc_ai_group @UNC, advised by. @mohitban47 | intern @googleCloud | ex RF @MSFTResearch
joykirat18.github.io
Joined April 2017
721
Following
467
Followers
RepliesMedia
  • Pinned
    user avatar
    Joykirat
    @joykiratsingh
    May 13
    🚨Excited to announce Agent-BRACE! LLM agents in long-horizon POMDPs either blow up their context with raw history or summarize it, discarding uncertainty by collapsing belief into a point estimate. Agent-BRACE decouples the agent into belief state + policy models, jointly
    Image
  • user avatar
    Joykirat
    @joykiratsingh
    May 20
    🚨 Excited to introduce MINTEval, a benchmark designed to evaluate memory system under heavy interference across evolving contexts and various challenging question types (long-range lookback & reasoning over multiple targets). Most memory benchmarks assume static, independent
    user avatar
    hyunji amy lee
    @hyunji_amy_lee
    May 20
    LLM agents & memory systems operate in continuously updated environments (Git repos, evolving docs). They must process long contexts, recover earlier information, and reason over many updates that create interference between old and new information. How well do they handle this?
    Image
  • user avatar
    Joykirat
    @joykiratsingh
    Apr 7
    🚨Excited to share Cog-DRIFT: When problems are too hard (pass@64=0), standard RLVR post-training like PPO and GRPO can fail: The model never finds a correct rollout, leading to zero reward. Cog-DRIFT reformulates hard open-ended problems into simple MCQ or cloze variants so
    user avatar
    Justin Chih-Yao Chen
    @cyjustinchen
    Apr 7
    🚨Cog-DRIFT: Breaking the Exploration Barrier in RLVR RLVR has pushed LLM reasoning forward BUT hits a ceiling: if a model can't solve a problem (rollouts never succeed), it gets 0 learning signal 👉Hard problems stay unsolved, and training stalls. We introduce✨Cog-DRIFT✨to
    Image
  • user avatar
    Joykirat
    @joykiratsingh
    Dec 17, 2025
    Huge congratulations to my advisor @mohitban47, on being elected an ACL Fellow! 🎉 This is truly well-deserved, and I’m deeply grateful for his mentorship, guidance and constant encouragement as his student.
    user avatar
    Mohit Bansal
    @mohitban47
    Dec 14, 2025
    Deeply happy and honored to be elected as an ACL Fellow -- and to be a part of the respected cohort of this+past years' fellows (congrats everyone)! 🙏 All the credit (and sincere gratitude) to all my amazing students, postdocs, collaborators, mentors, and family! 🤗💙
    Image
  • user avatar
    Joykirat
    @joykiratsingh
    Nov 4, 2025
    Thrilled to have our paper “Data-scarce Behavior Editing of Language Models” accepted at #EMNLPFindings2025! 🎉 We propose TaRot, a gradient-free method to edit LLM behavior efficiently — no retraining or large datasets needed. Super fun collab with Subhabrata Dutta &