Siddharth Ancha

@siddancha

Research Scientist @UCBerkeley • Postdoc @MIT_CSAIL @MIT • PhD from @SCSatCMU @CarnegieMellon • I work on Robotics 🤖

Berkeley, CA

siddancha.github.io

Replies Reposts Media

Pinned
Siddharth Ancha
@siddancha
Aug 21
Why is action chunking so important for robot behavior cloning? Even when open-loop execution reduces the policy’s reactivity? 🤖 Our latest work identifies the most important reason: **action chunking helps short-memory policies imitate long-memory experts**. More precisely,
Michael Zeng
@michzeng_
Aug 20
02:04
Action chunking — especially executing long action sequences open-loop — is widely used in imitation learning for robotic manipulation. Why is it so effective and do we really need it? We find a key reason: Long open-loop execution helps short-context policies imitate
1
Siddharth Ancha
@siddancha
Sep 15
Also, Dandelion Chocolate is hands down the best chocolate in the Bay area! 🍫🌉
Chelsea Finn
@chelseabfinn
Sep 15
A video from a Pi robot deployed at Dandelion Chocolate, fully autonomous w/ no interventions. 🤖 Deploying robots has taught us surprising lessons about the gap between proof-of-concept (i.e. building one box) and real-world utility (productively building boxes for hours). I
00:00
Siddharth Ancha
@siddancha
Sep 15
We live in interesting times! Previously we would painstakingly collect a labeled dataset for each task, train models and hope they would generalize (they wouldn’t). Now we can ask agents (with appropriate harnesses) to do just about anything. Token heavy but works zero shot!
Jitendra MALIK
@JitendraMalikCV
Sep 15
In the old days we used to talk about "graduate student descent" or "researcher gradient descent" as referring to the phase in research where one tries various techniques, changes parameters in the hope of getting to some desired output. Now we can get AI agents to do it. The
Siddharth Ancha
@siddancha
Jun 22
Really elegant work on fixing variational RL with diffusion models using self-distillation! 👏👏 *Problem*: Likelihood gradients ∇ log p(a) are needed for RL but are hard to compute for diffusion models. A useful surrogate is ∇ ELBO(a), where the ELBO is a supervised diffusion
Haozhe Jiang
@erichzjiang
Jun 16
Why aren’t Diffusion Language Model smart yet? Lacking stable post training is a major bottleneck! Meet DiPOD: the tripod for diffusion model post-training. DiPOD boosts accuracy across reasoning tasks, with Sudoku jumping from 22% to 97%, through a one-line code change. 🧵1/5
00:00
3
Siddharth Ancha
@siddancha
Jun 12
Cool work on refining coarse VLM actions using a flow matching policy π(a₀ | o) → a₁ where a₀ ∼ N(0, 1) by first reversing (inverting) the given coarse action a₁ via â₀ = π⁻¹(a₁ | o) and then reconstructing it in the forward direction i.e. â₁ = π(â₀ | o). The
Sergey Levine
@svlevine
Jun 12
Flow reversal steering allows "steering" diffusion-based VLAs with high-level actions, for example from VLM reasoning. This also lets us run RL in the diffusion noise space with exploration guided by high-level reasoning: think through a task, then practice it! 👇
00:00
1