- Rating
- 3
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
MONAD represents each continuous action vector as one latent token and reconstructs action blocks using a shared diffusion vision-language backbone with lightweight action interfaces. The approach is evaluated on LIBERO, LIBERO-Plus, and CALVIN, with additional ablations and temporal-response diagnostics.
Critical strengths
The unified architecture generates continuous actions without a separate action expert or pretrained action tokenizer. Competitive results across three simulation benchmarks, particularly LIBERO-Plus, demonstrate the practical promise of this design.
Critical weaknesses
- The benefit of the proposed reconstruction formulation is not sufficiently isolated. The paper lacks comparisons against direct regression and flow matching under the same backbone, data, and training budget, while the auxiliary-coordinate experiment in Section 4.4 does not directly establish the advantage of retaining the pretrained reconstruction formulation.
- The headline parameter accounting and temporal-response diagnostics come from legacy non-BARD models rather than the main benchmark checkpoints. Reporting the corresponding measurements for the evaluated BARD models would clarify how these architectural claims relate to the reported performance.
Questions for authors
Please address the concerns raised in the Weaknesses section.
3 comments
[W2, Q1] Reporting exact parameter accounting and finite-difference temporal-response (NLR) diagnostics on the evaluated
BARD-VL-4Bcheckpoints.
Thank you for catching this mismatch. Below we report exact parameter accounting and float32 finite-difference non-local temporal response (NLR) diagnostics measured directly on all four evaluated BARD-VL-4B (MONAD) checkpoints (Pretrain, LIBERO, CALVIN, and bimanual 20D RoboTwin). In the revised manuscript, we have updated the Abstract, Section 1, Section 4.6 (formerly Section 4.5), Section 5, and Appendix B.9 (Tables 14–15) by replacing the earlier 8.35B / 0.45% preliminary figures and NLR table with these BARD-VL-4B measurements so that all architectural and mechanistic statistics align directly with the main benchmark checkpoints.
BARD-VL-4B (MONAD) Module | Parameters | % of Total (4.46B) | Architectural Specification |
|---|---|---|---|
| Shared VLM Backbone | 4,437.82M | 99.48% | 24-layer ViT (415.35M) + 36-layer LM (4,022.47M, , tied language head) |
| Separate Action Transformer | 0 | 0.00% | None (vs. 14.3%–51.5% in , , X-VLA, GR00T-N1.7) |
| Action/State Bridge & Risk Modules | 23.26M (10.15M active) | 0.52% (0.23% active) | Point-wise 4-layer state/action MLPs (9.49M) + 2-layer token-risk head (0.66M) |
| Total Parameters | 4,461.08M | 100.00% | Unified BARD-VL-4B parameter count across all evaluated checkpoints |
Following Eq. (5) of the paper, we also evaluated central finite differences ( in exact float32, averaged over 11 deterministic probes) across all four BARD-VL-4B checkpoints under both minimal (lang + state) and real dataset observations:
BARD-VL-4B Checkpoint | Context & Mask (, ) | Point-Wise Bridge NLR | Backbone Layer (L0 -> L9 -> L18 -> L27 -> L35) | Output | Output | Output |
|---|---|---|---|---|---|---|
| Pretrain | Minimal, Bidirectional (20, 20D) | 0.00% | 2.53 -> 3.58 -> 15.46 -> 29.11 -> 34.79% | 36.12% | 10.06% | 4.23% |
| Pretrain | Matched, Bidirectional (50, 20D) | 0.00% | 1.79 -> 2.32 -> 9.95 -> 24.65 -> 28.55% | 30.68% | 9.13% | 7.13% |
| LIBERO | Real Obs, Bidirectional (50, 10D) | 0.00% | 0.34 -> 0.87 -> 9.00 -> 23.46 -> 29.99% | 23.52% | 2.55% | 0.84% |
| LIBERO | Real Obs, Block-Causal (50, 10D) | 0.00% | 0.37 -> 0.96 -> 9.43 -> 24.77 -> 30.34% | 22.49% | 2.02% | 0.63% |
| CALVIN | Real Obs, Block-Causal (25, 10D) | 0.00% | 2.66 -> 3.01 -> 11.77 -> 28.12 -> 44.74% | 37.44% | 5.04% | 1.75% |
| RoboTwin | Minimal, Block-Causal (50, 20D) | 0.00% | 2.29 -> 2.66 -> 10.97 -> 29.37 -> 36.36% | 34.79% | 4.18% | 1.34% |
These measurements confirm that our architectural and mechanistic claims hold directly on the evaluated BARD-VL-4B models. Across all 19 float32 configurations (and step-size sweeps , where and vary by < 0.01 percentage points with zero repeat-forward jitter), the point-wise 4-layer MLP bridge exhibits strictly 0.00% cross-timestep coupling, whereas a single pass through the shared VLM backbone routes 20.24%–38.05% () and 0.51%–7.13% (, matching the 1.44%–7.14% scale reported for our preliminary 8B models) of perturbation energy across distinct action timesteps, growing progressively from shallow (0.34%–4.91% at Layer 0) to deep shared layers (21.09%–46.96% at Layers 27–35).
Thank you for the detailed rebuttal and the revised manuscript; the matched-budget comparison against direct regression and flow matching and the BARD-VL-4B parameter and NLR measurements have fully addressed my concerns, so I will maintain my original score.
Thank you for this valuable suggestion. We conducted a strictly controlled closed-loop and offline benchmark on the 10 long-horizon multi-stage tasks of
LIBERO-10using the exact same4.46BsharedBARD-VL-4B Pretrainbackbone, identicalLIBERO-10training stream, identical trainable parameter count (1.23B[1,234,432,533] trainable parameters: top 12 shared VLM layers + identical 4-layer point-wise I/O bridge and risk head), and identical 2,000-step budget (effective_batch_size = 32,64,000trajectory chunks,31.63 GiBgradient-checkpointed peak VRAM, FP32 master-weight AdamW,lr = 2.5e-5). Across 100 paired closed-loop episodes (10tasks ×10paired seeds per arm) and 64 held-out trajectory windows, we compared MONAD (Masked Reconstruction) (self-conditioned draft-and-remask reconstruction over[MASK]action slots, evaluated under both single-passD1/R0and risk-guidedD4/R2), Shared + Direct Regression (single-pass L1/BCE regression from 100%-masked query slots), and Shared + Flow Matching (conditional optimal-transport velocity matching v=x1−z0 from z0∼N(0,I), evaluated with 6-step Euler integration):4.46BShared Backbone,1.23BTrainable)LIBERO-10Closed-Loop Success (100Paired Eps) ↑D4/R2)D4/R2(Risk-Guided)D1/R0(1-Pass Draft)-2.0 pp-14.0 pp(95% CI:[-23.0, -5.0])-70.0 pp(95% CI:[-80.0, -59.0])First, comparing MONAD against Flow Matching directly establishes the advantage of retaining the pretrained masked-reconstruction formulation. Adapting a discrete
[MASK]-pretrained VLM backbone with continuous Gaussian interpolants zt=(1−t)z0+tx1 and velocity targets x1−z0 creates a severe representation mismatch during downstream adaptation—degrading action MAE (0.447at 6 Euler steps;0.784at 1 step), gripper accuracy (80.0%vs.92.8%), conditional 50% completion MAE (0.195vs.0.055), and paired closed-loop success onLIBERO-10by70.0percentage points (21.0%vs.91.0%, paired 95% bootstrap CI[+59.0%, +80.0%]). Retaining discrete masked reconstruction preserves seamless alignment with masked multimodal-action pretraining.Second, comparing MONAD against Direct Regression highlights why closed-loop physical execution and conditional trajectory coupling matter far more than unconditional open-loop L1 error alone. Because
Direct Regressiondedicates 100% of its gradient updates to the fully masked query state (t=1), it attains slightly lower unconditional open-loop MAE (0.316vs.0.339); however, supervising only t=1 sacrifices intra-chunk conditional trajectory coupling—eroding 50% trajectory completion MAE from0.056at initialization to0.326(6.0×higher error than MONAD's0.055)—and restricts the policy to a deterministic single-pass average across multi-modal contact transitions. By contrast, MONAD's masked-reconstruction objective jointly trains full-mask drafting and partial-mask conditional denoising, regularizing the shared backbone to capture coherent cross-timestep dependencies: even at a single forward pass (D1/R0), MONAD outperformsDirect Regressionby+12.0percentage points in pairedLIBERO-10closed-loop rollouts (89.0%vs.77.0%), and risk-guided iterative remasking (D4/R2) further lifts closed-loop success to91.0%(+14.0percentage points, paired 95% CI[+5.0%, +23.0%]) (92.0%withD4/R4on our full multi-suite checkpoint in Reviewer tnFY [W2]). In the revised manuscript, we have added this parameter- and budget-matched benchmark against Direct Regression and Flow Matching—together with the analysis of pretraining paradigm alignment and conditional trajectory coupling—to Section 4.5 (Table 2).