- Rating
- 3
- Confidence
- 4
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 3
Read review
Paper summary
This paper proposes Divide and Prune (DiP), a training-free visual-token pruning method for driving VLAs. DiP uses Depth Anything v2 to divide image tokens into near- and far-field groups, retains near-field tokens through a rank fusion of Sobel edge energy and token -norm, and selects far-field tokens through depth-modulated cosine farthest-point sampling. On Impromptu-VLA/nuScenes, DiP is compared with SparseVLM, VisPruner, DivPrune, and FastDriveVLA at several token budgets; on SimLingo/Bench2Drive, it evaluates aggressive pruning in closed loop. At 75% token pruning, the method reports an average L2 error of 31.97 cm and reduces LLM prefill time from 185 ms to 55 ms.
Critical strengths
-
DiP provides a practical accuracy-efficiency trade-off. With 812 of 3,249 tokens retained, it preserves an average L2 error of 31.97 cm while reducing reported prefill time from 185 ms to 55 ms.
-
The evaluation spans two VLA backbones and closed-loop driving. DiP retains 96.0% of the full model's Driving Score on Bench2Drive with only 25% of the visual tokens.
Critical weaknesses
-
The published far-field algorithm is inconsistent with the stated method. is never used in Algorithm 1, and all similarity matrices and selections are defined over , which includes valid near-field tokens. The paper must clarify whether the implementation instead uses and how overlap with is prevented (Section 3.4, Equations 7–8, p. 5; Algorithm 1, p. 14).
-
The closed-loop and safety claims are stronger than the evidence. Table 2 compares DiP with heterogeneous full-token systems rather than pruning methods applied to the same SimLingo checkpoint, and the pruned model loses substantial Success Rate and Comfortness relative to SimLingo. Figure 5 shows qualitatively better token coverage but does not by itself establish safer control. In addition, the reported 3.36× speedup measures LLM prefill rather than complete action-generation or control-loop latency (Table 2 and Figure 5, pp. 7–9; Table 5, p. 8).
Questions for authors
-
Which Table 2 baselines were reproduced and which were quoted from prior work? For methods with available implementations, can the authors provide at least one same-SimLingo, same-token-budget pruning comparison to contextualize DiP's closed-loop result (Table 2, p. 7; Appendix D, p. 15)?
-
Can the authors provide a compact equal-budget ablation that separately removes the depth partition, the near-field energy/norm selector, and the far-field depth modulation? These three variants would be sufficient to identify the source of the gain without requiring an exhaustive factorial study (Sections 3.2–3.4 and Section 4.3, pp. 4–8).
-
How do the authors reconcile the repeated safer-driving interpretation with the lower Success Rate and Comfortness and the slightly worse full-token collision/intersection rates under aggressive pruning (Tables 1–2, p. 7; Figure 5, p. 9)?