Research @OpenAI. I study Reinforcement Learning. PhD from UT Austin. Previously FAIR Paris, Meta US, NVIDIA, CMU, and IIT Kharagpur. Website: https://hari-sikchi.github.io/https://bsky.app/profile/harshitsikchi.bsky.socialhttps://bsky.app/profile/harshitsikchi.bsky.social/post/3lcz7aveq6k26🤖 Introducing RL Zero 🤖: a new approach to transform language into behavior zero-shot for embodied agents without labeled datasets!11 Dec 2024 07:11 +0000at://did:plc:ahnmadduif6wtgafkqs6payf/app.bsky.feed.post/3lcz7aveq6k26https://bsky.app/profile/harshitsikchi.bsky.social/post/3lcefcjz4ec2pWhat if I told you all solutions for RL lie on a (hyper) plane? Then, we can use that fact to learn a compressed representation for MDP that unlocks efficient policy inference for any reward fn. On this plane, solving RL is equivalent to solving a linear constrained optimization!03 Dec 2024 00:33 +0000at://did:plc:ahnmadduif6wtgafkqs6payf/app.bsky.feed.post/3lcefcjz4ec2phttps://bsky.app/profile/harshitsikchi.bsky.social/post/3lcbv6wgtkk2pI will be attending @neuripsconf.bsky.social and am on the job market. Hit me up to chat about topics in RL (Zero-shot RL, Imitation Learning, Offline RL, Deep RL) or Alignment! Learn more about my research interests: hari-sikchi.github.io/research/ https://hari-sikchi.github.io/research/02 Dec 2024 00:39 +0000at://did:plc:ahnmadduif6wtgafkqs6payf/app.bsky.feed.post/3lcbv6wgtkk2phttps://bsky.app/profile/harshitsikchi.bsky.social/post/3lbrxcx4z222sThis is just a bad year for ICLR authors and reviewers 😥25 Nov 2024 16:35 +0000at://did:plc:ahnmadduif6wtgafkqs6payf/app.bsky.feed.post/3lbrxcx4z222s