Skip to content

Ep#105: Beyond Imitation: Self-Improving Robot Policies via Off-Policy Q-Planning

September 18, 2026 · 50 min

Show notes

Imitation learning, especially with interventions, has driven so much recent robotics progress. However, improving a policy via targeted interventions until it reaches a useful and deployable success rate is a time and labor intensive process. Instead, wouldn’t it be great if policies could improve on their own?

That’s what Varun Giridhar and Animesh Garg join us to talk about. In Q-Planning, they start with a large policy like pi-0.5, and add a Q-function estimator to predict value instead of just actions, then use both successful and failed rollouts to update this Q-function online, then use it to guide sampling and trajectory selection. With just a few rollouts they can dramatically improve policy performance online. This provides a way to do really difficult tasks like inserting a credit card into a wallet, increasing success rate from 25% to 80% in just a few iterations.

Learn more in Episode 105 of RoboPapers, hosted by Michael Cho, Chris Paxton, and Ruijie He!

Abstract

Behaviour Cloning (BC) has driven remarkable progress in robot manipulation, yet it is fundamentally limited by its inability to self-improve: a policy that fails cannot learn from that failure without additional human demonstrations. Reinforcement Learning fine-tuning offers a path to self-improvement but has proven difficult to scale to the multi-billion-parameter models underpinning modern robot policies. We propose Q-Planning, which equips a large visuomotor BC policy with a small off-policy Q-function. Because a Q-function estimates value rather than imitates actions, it can be trained on the same successful demonstrations as the BC policy and later absorb both successful and failed deployment rollouts, an asymmetry BC does not have. We exploit this asymmetry to enable value-guided action selection at inference (a single-step QQ-weighted average over BC draws) and online self-improvement that fine-tunes only the Q-function, leaving the BC weights untouched. On LIBERO and bimanual RoboTwin, ten iterations of self-improvement lift every benchmark score we tested (LIBERO-10 93 → 99%, RoboTwin 83.8 → 91.4%) and shorten successful episodes on the near-ceiling suites (LIBERO-Object, LIBERO-Goal). On two contact-rich bimanual real-robot tasks, the same loop (BC frozen, no human intervention) improves purely from its own deployment rollouts: stack-cups 40 → 90% and insert-wallet 25 → 80% in five iterations, whereas SFT on successful rollouts alone stalls at 55% and 30%. Under an identical online budget Q-Planning is the only method, among Best-of-N, filtered SFT, IBRL, DSRL, and DAWR, that improves stably from failures without training an auxiliary actor.

Learn More

Project page: https://q-planning.github.io/

ArXiV: https://arxiv.org/abs/2608.21204



This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit robopapers.substack.com

All episodes of RoboPapers

Settings

Display

Language

Decides which stations you see first on the home page.

History

Recently played stations and episodes and your latest searches are only stored in this browser.

Notifications

A short notice when a podcast you follow has a new episode. Turn off single podcasts with the bell on their page.

Account

Save favorites – in the app and on the website.

Keyboard shortcuts

/
Open search
Space
Play or pause
← →
In the player: back 15 s, forward 30 s (episodes) · in a row: next tile
↑ ↓
In a row: target within the tile
Home / End
In a row: first or last tile
Esc
Close window or menu
?
This overview
Now playing

Go to station page

Play next

    Add episodes with “Play next” – they play after the current episode.