- Rating
- 2
- Confidence
- 4
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
The paper proposes ExAbs, which is a method for learning execution-aware temporal abstractions in hierarchical world models. Standard hierarchical models learn high-level waypoints from offline trajectories without considering whether the low-level controller can actually reach them within budget. ExAbs addresses this with two components: an Achievability Model (AM) that learns from closed-loop execution the probability a waypoint is reachable within a given budget and uses this to select segment boundaries and score candidate waypoints during planning; and PETAL, which distills a frozen planner’s decisions into compact source-relative effect representations so local action selection compares requested versus predicted changes rather than endpoint proximity alone. Collecting AM labels with the PETAL-enabled controller ties learned abstractions to the actual executor. Across seven simulated tasks on four world-model backbones and three real-robot tasks,
Critical strengths
1. AM innovatively uses achievability to judge the quality of candidate transitions: it asks whether the controller can actually complete a transition within a given budget, which rules out the "predictable but infeasible" subgoals that arise in conventional approaches.
2. AM and PETAL are not isolated modules; the authors organise them into a coherent whole. The implementation is careful as well: PETAL is frozen before AM labels are collected, and AM is frozen before the boundaries are selected, which avoids the problem that a changed local objective invalidates previously collected labels. I particularly appreciate this.
3. The paper is well written.
Critical weaknesses
1. PETAL's motivation is unclear. The authors state that "full-state similarity is shaped largely by content that the actions do not change", and §5.2 / Appendix K.2 further gloss this as "appearance changes that leave the physical transition unchanged". Unless I misunderstand the setup, every experiment in the paper uses a fixed-view camera, and on the physical robot even the scene fixtures remain fixed. Appearance should therefore not change within an episode; what changes is the part related to the agent's pose. Moreover, since the requested waypoint and the candidate endpoint both derive from the same , the content they share cancels in the difference . What can actually differentiate candidates is then only candidate-correlated prediction error, together with the mismatch between the retrained high-level predictor (which produces ) and the frozen low-level predictor (which produces ). It is not the appearance that the authors emphasise.
2. PETAL does contribute, but no internal ablation is provided, and the distribution of the gains points to another mechanism. The improvement of +PETAL over L2 is real (Table 5: Reach-Wall 36.5→47.9, Counterflow 37.5→56.5, Two-room 9.4→15.6; Table S11: PushT , 28.7→32.7), yet as far as I can see the paper does not ablate PETAL's internal design. The distribution of the gains also points to a different mechanism: +PETAL is largest exactly on the tasks where the agent must temporarily move away from the waypoint (Reach-Wall +11.5, Counterflow +19.0, Two-room +6.3) and small where it need not (Maze-Hard +2.5; PushT +2 to +4). This is what §3.3 describes when it says that the anchored terminal comparison replaces the original path-summed goal cost: a path-summed cost toward a single waypoint penalises precisely those necessary detours, whereas appearance invariance would not concentrate the gains on obstacle and constraint tasks. I therefore suspect that what actually drives PETAL is not the part the authors advertise.
Questions for authors
1. What is the observation setup for the simulated tasks: a single world-fixed camera, or an egocentric view? Is the camera pose constant within an episode? Are goal images rendered under the same camera, lighting and materials?
2. Which part of PETAL produces the gain, and does the paper provide the corresponding ablation (for example , i.e. the anchor alone; a fixed metric instead of the learned one; the support mask disabled; the top-128 shortlist removed; and terminal versus path-summed comparison)?
7 comments
[Part 2 of 3 — Q2 & W2: Endpoint Cost vs. Learned Effect Representation]
[Q2 & W2 — Part A] Endpoint Scoring (T0/T1) vs. Learned Effect Encoder (A/R/P)
Thank you for the concrete ablation list. We ran all requested controls under matched world models, queries, budgets, and planning seeds with online AM disabled throughout (Section 5.3, Appendix L.4, Tables S17–S19). In short: switching from path-aggregated to endpoint-only base cost changes success by to pp, whereas holding the anchor fixed and adding the learned effect encoder adds to pp across all seven tasks (+9.1 pp pooled over 560 episodes).
1. Path-aggregated vs. endpoint-only base cost without PETAL (T0 vs. T1, Table S17). Reach-Wall, PushT, Cube, Button–Window, and Button–Drawer already evaluate endpoint error in their native cost or anchor (Section 3.3, Appendix D.2), so replacing path aggregation cannot explain +PETAL's +11.5 pp gain on Reach-Wall. Path aggregation is replaced only in Maze (path sum), Two-room (path mean), and Counterflow (final 5-step mean). Testing native path cost (T0) against endpoint-only distance (T1) without PETAL never improves success by more than +1.2 pp and lowers the other three settings:
| Task | Queries | T0 (Native Path Cost) | T1 (Endpoint-Only Cost) | T1 T0 (pp) |
|---|---|---|---|---|
| Counterflow | 200 | 40.67% | 41.83% | +1.17 |
| Two-room | 64 | 10.94% | 7.81% | -3.13 |
| Maze Medium | 40 | 95.00% | 92.50% | -2.50 |
| Maze Hard | 40 | 82.50% | 80.00% | -2.50 |
2. Anchor Alone vs. Random Encoder vs. Learned PETAL (A/R/P, Table S18). Holding the complete PETAL anchor (including its temporal and spatial form) and single-seed evaluation protocol fixed across arms, we compare:
- A (): PETAL's anchor alone, with the effect branch disabled.
- R: Identically structured random encoder frozen at initialization, with its comparison network retrained from scratch on the same teacher data, budget, and validation rule (isolating Eq. 5 + Eq. D4–D5 supervision of ).
- P: Trained PETAL encoder and comparison network.
| Task | Queries | A (, Anchor) | R (Random ) | P (Learned PETAL) | P A (pp) | P R (pp) |
|---|---|---|---|---|---|---|
| Reach-Wall | 96 | 36.46% | 42.71% | 47.92% | +11.46 | +5.21 |
| Counterflow | 200 | 42.50% | 40.50% | 54.50% | +12.00 | +14.00 |
| Button–Window | 60 | 3.33% | 0.00% | 18.33% | +15.00 | +18.33 |
| Two-room | 64 | 10.94% | 4.69% | 15.63% | +4.69 | +10.94 |
| PushT | 50 | 28.00% | 22.00% | 32.00% | +4.00 | +10.00 |
| Maze Hard | 40 | 85.00% | 80.00% | 87.50% | +2.50 | +7.50 |
| Cube | 50 | 60.00% | 56.00% | 62.00% | +2.00 | +6.00 |
| Total (560) | 560 | 36.96% (207) | 35.00% (196) | 46.07% (258) | +9.11 | +11.07 |
With the anchor held fixed, learned effects (P A) add +11.5 pp on Reach-Wall, +12.0 pp on Counterflow, +15.0 pp on Button–Window, and +4.7 pp on Two-room. P also exceeds R in all 7 tasks (+5.2 to +18.3 pp), showing that the gain comes from learning rather than adding a trainable scorer. Obstacle and precondition tasks benefit most because over a short 10–20 step local horizon, candidates stalling near a barrier and candidates initiating a tangential detour or button press can have similar endpoint distance , whereas directional source-relative effects separate clearance motion from stalled rollouts.
[Part 3 of 3 — Q2 & W2: Metric, Spatial Support & Shortlist Ablations]
[Q2 & W2 — Part B] Remaining Design Choices in Table S19
Below we examine three further design choices in Appendix L.4 and Table S19 under matched evaluation protocols with online AM disabled throughout. Each ablation covers every setting whose adapter includes that component.
FixedD retains the learned effect encoder and replaces the learned comparison metric with squared Euclidean distance, testing the contribution of metric learning. Uniform replaces selective spatial support with uniform spatial weights, testing the contribution of spatial selection. NoSL scores the full candidate pool instead of the top-128 shortlist, testing whether candidate filtering materially affects success.
3. Learned Metric vs. Squared Euclidean Distance Across All 9 Settings (Table S19a):
| Task | FixedD (Fixed ) | P (Learned Metric) | P FixedD (pp) |
|---|---|---|---|
| Maze Medium | 92.50% | 95.00% | +2.50 |
| Maze Hard | 85.00% | 87.50% | +2.50 |
| PushT | 28.00% | 30.00% | +2.00 |
| Two-room | 14.06% | 15.63% | +1.56 |
| Reach-Wall | 46.88% | 47.92% | +1.04 |
| Counterflow | 56.17% | 56.17% | 0.00 |
| PushT | 18.67% | 18.67% | 0.00 |
| Button–Window | 18.33% | 18.33% | 0.00 |
| Cube | 64.00% | 62.00% | -2.00 |
4. Selective Spatial Support vs. Uniform Pooling Across All 6 Spatial-Grid Settings (Table S19b): LeWM uses 1D vector latents on PushT and Cube without a spatial feature grid. Across all six tasks with spatial feature maps, selective support is strictly higher in 5 settings and ties at 95.0% on Maze Medium:
| Spatial-Grid Task | Uniform (No ) | P (Selective ) | P Uniform (pp) |
|---|---|---|---|
| Reach-Wall | 44.79% | 47.92% | +3.13 |
| Counterflow | 53.67% | 56.17% | +2.50 |
| Maze Hard | 85.00% | 87.50% | +2.50 |
| Button–Window | 16.67% | 18.33% | +1.67 |
| Two-room | 14.06% | 15.63% | +1.56 |
| Maze Medium | 95.00% | 95.00% | 0.00 |
5. Top-128 Candidate Shortlist vs. Full-Pool Scoring Across All 4 Shortlisted Settings (Table S19c): The remaining adapters already score all candidates without truncation. Removing the top-128 shortlist changes success by to pp, indicating it serves as a compute optimization rather than the source of PETAL's gains:
| Shortlisted Task | NoSL (Full Pool) | P (Top-128 Shortlist) | P NoSL (pp) |
|---|---|---|---|
| Maze Medium | 92.50% | 95.00% | +2.50 |
| Maze Hard | 87.50% | 87.50% | 0.00 |
| Counterflow | 57.33% | 56.17% | -1.17 |
| Button–Window | 20.00% | 18.33% | -1.67 |
These ablations show that metric learning and selective spatial support can improve task success, while the shortlist has small effects that vary across tasks. The learned metric matches or exceeds FixedD in 8 of 9 settings (strictly higher in 5), and selective spatial support matches or exceeds Uniform in all 6 spatial settings (strictly higher in 5). Since the success-rate changes from these three components are limited to about 3.1 pp, they can be viewed as secondary design choices. Taken together, Tables S17–S19 suggest that PETAL's gains arise primarily from learning the source-relative effect representation .
[Part 1 of 3 — Q1 & W1: Observation Setup & Static-Background Analysis]
Dear Reviewer X9n9,
Thank you for the careful reading and for the precise questions about PETAL. Part 1 addresses Q1 and W1, Part 2 covers endpoint scoring (T0/T1) and the learned effect encoder (A/R/P), and Part 3 reports the remaining internal ablations (FixedD, Uniform, NoSL). All updates appear in the uploaded revision.
[Q1 & W1] Camera Setup and Why Static Backgrounds Do Not Cancel Exactly
1. Answer to Q1 (Appendix H.1). All seven primary simulated tasks (plus Button–Drawer) use a single world-fixed camera rather than an egocentric view, with constant camera pose within and across episodes: global 2D renderings in Counterflow and Two-room, fixed top-down views in Maze and PushT, and fixed external views in Cube, Reach-Wall, Button–Window, and Button–Drawer (supplementary Visual AntMaze follows OGBench's default agent-tracking third-person camera). Goal images are rendered by the same simulator with the exact same camera, lighting, and materials as episode observations, while intermediate hierarchical waypoints are model-predicted latent states.
2. Response to W1 (Additive cancellation vs. nonlinear entanglement). We agree with your decomposition: if the encoder E mapped static background to an additive latent offset shared across candidates, that offset would cancel in z^i−w, leaving only (a) candidate-correlated FL prediction error and (b) FH-vs.-FL horizon/predictor mismatch. In practice, a third term enters: (c) learned ViT and ConvNet encoders are nonlinear and entangle static background context with agent and object pose, so background is not an additive constant even when camera and scene appearance stay fixed.
We test (c) directly with Control 1 (Real outcomes) and examine stress-test robustness with Controls 2–4 (Section 5.2, Appendix K.4, Table S16), holding camera geometry, physical starts/goals, dynamics, models, and budgets fixed with online AM disabled:
Across these five tasks, PETAL improves attainment under the changed goal background in both controls, suggesting benefits for local execution toward both expert and handoff targets. Meanwhile, native unshifted closed-loop gains are isolated directly in Part 2 below.