Rethinking Demonstration Unlearning in Imitation Learning for Robotics

• URL: http://arxiv.org/abs/2608.20784v1
• Date: Fri, 21 Aug 2026 06:53:55 GMT
• Title: Rethinking Demonstration Unlearning in Imitation Learning for Robotics
• Abstract summary: Metrics inherited from machine unlearning, such as forgetting loss or a single membership attack, do not establish what an edit removed from a policy acting in closed loop.<n>We therefore introduce a retrain-calibrated audit that reads demonstration unlearning along two axes: behavior, whether the edited policy acts like one retrained without the removed demonstrations, and evidence, whether an auditor can still detect it was trained on them.
• Score: 4.7767476304018075
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Imitation learning for robotics depends on human demonstrations, some of which people may later ask to remove. Retraining without them is the natural reference, but its cost grows with policy and dataset scale, motivating cheaper operators that edit a trained policy. Metrics inherited from machine unlearning, such as forgetting loss or a single membership attack, do not establish what an edit removed from a policy acting in closed loop. We therefore introduce a retrain-calibrated audit that reads demonstration unlearning along two axes: behavior, whether the edited policy acts like one retrained without the removed demonstrations, and evidence, whether an auditor can still detect it was trained on them. The behavior axis measures action divergence to that retrain at matched states, calibrated by a floor built from independent retrains, so a policy at the floor is as close to a retrain as retrains are to each other. The evidence axis applies a per-demonstration membership attack against a retrain null, reporting both its rank and its absolute member-loss level, since rank alone accepts operators that inflate member losses past the null. A conformal test then combines both axes into one hypothesis of joint retrain consistency, against a fleet of independent retrains large enough to reject at conventional significance. Across five preregistered conditions on three real-robot policy classes and two simulation suites, the axes dissociate in both directions on one checkpoint, as an edit may repair task behavior while leaving evidence unchanged, or reduce evidence while moving behavior away from retraining. On the ACT arm, a redirect edit restores blind-scored robot success to 18 of 20 trials.

Related papers

• The Latent That Never Was: A Forensic Re-run of the CVAE Ablation in Action Chunking Transformer [3.4858077573471102]
  Action Chunking Transformers (ACT) are widely used to learn robot manipulation from demonstrations.<n>Their conditional variational autoencoder includes an encoder meant to capture differences between demonstrations during training.<n>The original ACT paper reported that encoder removal dropped the mean success rate from 35% to 2% on two simulated tasks with human demonstrations.<n>We re-ran this ablation in the original code and checked whether the findings depend on the implementation or training data.
  arXiv  Detail & Related papers  (2026-09-15T07:18:26Z)
• RecurSE: Bounded Recursive Self-Evaluation for LLM Rubric Judges [48.831446371402414]
  RecurSE pairs a trainable judge with a synchronized policy-copy checker that audits the judge's reasoning against meta-rubrics to supply a scalar process reward.<n>We show that RecurSE achieves consistent generalization gains across held-out medical, pairwise, summarization, and professional benchmarks.
  arXiv  Detail & Related papers  (2026-08-25T08:35:04Z)
• Benign on Label, Malicious by Design: Clean-Label Dormant-to-Activated Backdoor via Machine Unlearning with Removable Camouflage [10.10240378510637]
  Existing backdoor attacks often become effective immediately after backdoor implantation and may therefore be exposed before exploitation.<n>Machine unlearning activated dormant backdoors by remaining inactive after training and becoming effective only after selected training records are unlearned.<n>We propose a clean-label unlearning-activated backdoor framework based on dual-generator learning and formulate it as a bilevel optimization problem.
  arXiv  Detail & Related papers  (2026-07-30T09:46:02Z)
• TOUR: A Trajectory-Level Unlearning Benchmark for Offline Reinforcement Learning [18.411026872661083]
  We introduce Trajectory-level memOrization and Unlearning in offline RL (TOUR)<n>TOUR combines trajectory-level partitioning, matched non-member controls, retraining references, retained-performance anchors, and privacy auditing.<n>We show that common deletion baselines have environment-dependent privacy-utility behavior.
  arXiv  Detail & Related papers  (2026-07-23T09:40:40Z)
• Zero2Skill: Bootstrapping Robot Skills through Autonomous Data Collection, Training, and Deployment [53.913175773174636]
  Existing pipelines reduce human effort via self-resetting, VLM verification, or language-guided correction.<n>We present Zero2Skill, a human-robot symbiotic agentic system in which corrections are retained and reused across rounds.
  arXiv  Detail & Related papers  (2026-07-15T17:16:24Z)
• Perfect Demo Makes Poor Teacher: Learning Robust Alignment from Critical Motion Segments [53.85970147743299]
  Expert demonstrations are widely assumed to be the gold standard for robot imitation learning.<n>Yet for fine-grained manipulation such as insertion, stacking, and alignment, we uncover a counterintuitive failure mode: fluent demonstrations can be poor teachers.<n>We introduce STAIR, a compact dynamic feature that bridges the vision-language model and the action expert, distilling the short-horizon motion already recorded in each trajectory into dense, motion-aware supervision.
  arXiv  Detail & Related papers  (2026-06-14T04:15:29Z)
• Survive or Collapse: The Asymmetric Roles of Data Gating and Reward Grounding in Self-Play RL [76.45061154544568]
  Self-play reinforcement learning trains language models on their own generated tasks, co-evolving a proposer and solver without human labels.<n>Recent systems report strong reasoning gains, but collapse and instability are widely observed and poorly understood.<n>We argue instead that self-play stability is governed by two distinct levers: a data-level gate that decides which proposer-generated tasks enter the training pool, and the reward signal that updates the policy on tasks already admitted.
  arXiv  Detail & Related papers  (2026-05-21T09:19:23Z)
• Conservative Offline Robot Policy Learning via Posterior-Transition Reweighting [36.23660195981511]
  Posterior-Transition Reweighting (PTR) is a reward-free and conservative post-training method.<n>It decides how much each training sample should influence the supervised update.<n>PTR reallocates credit according to how attributable each sample's post-action consequence is under the current representation.
  arXiv  Detail & Related papers  (2026-03-17T14:05:35Z)
• Embedding Classical Balance Control Principles in Reinforcement Learning for Humanoid Recovery [1.0883174135300417]
  Humans are vulnerable to falls and unrecoverable failure states, limiting their practical utility in unstructured environments.<n>We present a unified RL policy that addresses this limitation by embedding classical balance metrics.<n>The policy achieves a 93.4% recovery rate across randomized initial poses and unscripted fall configurations.
  arXiv  Detail & Related papers  (2026-03-09T17:02:30Z)
• A Granular Study of Safety Pretraining under Model Abliteration [64.24346997570275]
  We study model abliteration, a lightweight projection technique designed to remove refusal-sensitive directions.<n>We issue 100 prompts with balanced harmful and harmless cases, classify responses as **Refusal** or **Non-Refusal** using multiple judges, and validate judge fidelity.<n>Our study produces a checkpoint-level characterization of which data-centric safety components remain robust under abliteration.
  arXiv  Detail & Related papers  (2025-10-03T07:01:45Z)
• Reinforcement Learning with Verifiable yet Noisy Rewards under Imperfect Verifiers [90.50039419576807]
  Reinforcement Learning with Verifiable Rewards (RLVR) trains policies against automated verifiers to avoid costly human labeling.<n>To reduce vulnerability to verifier hacking, many RLVR systems collapse rewards to binary $0,1$ during training.<n>This choice carries a cost: it introduces textitfalse negatives (rejecting correct answers, FNs) and textitfalse positives (accepting incorrect ones, FPs)
  arXiv  Detail & Related papers  (2025-10-01T13:56:44Z)
• Rethinking Classifier Re-Training in Long-Tailed Recognition: A Simple Logits Retargeting Approach [102.0769560460338]
  We develop a simple logits approach (LORT) without the requirement of prior knowledge of the number of samples per class. Our method achieves state-of-the-art performance on various imbalanced datasets, including CIFAR100-LT, ImageNet-LT, and iNaturalist 2018.
  arXiv  Detail & Related papers  (2024-03-01T03:27:08Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.