Darsh Khandelwal

[Darsh Khandelwal @DarshK_2023]

@jaswwei Very very interesting, so cool!!

[Darsh Khandelwal @DarshK_2023]

@tanaytndn @ScaleAILabs Impressive work, results look very promising!Congrats on NeurIPS Workshop acceptance, looking forward to seeing what work builds on this.

[Tanay Tandon @tanaytndn]

Over the summer at @ScaleAILabs, I investigated whether synthetic data augmentation can help robots generalize.

With the paper now accepted to the NeurIPS 2026 Robot Learning Workshop, I wanted to share how we approached this and what we found.

[Kai Zou @zkjzou]

Not All Privileged Information Is Equal: What Makes OPSD Actually Work?

We find that information-rich PI, such as environment observations and mistake analysis, outperforms PI that only provides ground-truth solutions.

Better PI leads to better post-training performance. But can we predict which PI will work best before spending compute on post-training?

We investigated three offline metrics:
🔹 Utility: A model's performance with PI can be misleading due to information leakage.

🔹 Reverse KL: Larger PI-induced distribution shifts often correlate with better post-training performance, but incorrect PI can cause equally large shifts.

🔹 Error-step KL contrast: Measuring whether PI changes the model's behavior specifically at error steps better captures its effectiveness and helps select stronger PI.

What types of privileged information have you found most effective in your OPSD experiments?

📄 Paper: https://t.co/kNQO1RD1T6

[Darsh Khandelwal @DarshK_2023]

@estellespace Is that the dim sum leftovers? 😭

[Darsh Khandelwal @DarshK_2023]

@tanaytndn Ik

[Darsh Khandelwal @DarshK_2023]

@jaswwei Insightful ☝️