Unified Prediction and Planning via Conflict-Aware Disjoint Parameter Training

Daehee Park; Giwon Lee; Kuk-Jin Yoon; Seonae Jeon; TaeWon Seo

REVIEW 4 major objections 5 minor 17 references

Unified Prediction and Planning via Conflict-Aware Disjoint Parameter Training

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read In compact unified models, prediction and planning compete for the same weights; the paper shows that training them on disjoint parameter regions and then sparsely merging the top 1% of each task's changes resolves the conflict and roughly

desk verdict Solid, well-documented empirical method for compact prediction/planning; the 'skill conflict' story needs a random-mask control before I'd fully buy it. read the letter →

arxiv 2607.19971 v1 pith:P56HPZOS submitted 2026-07-22 cs.RO cs.CV

Taewon Seo , Seonae Jeon , Giwon Lee , Kuk-Jin Yoon , Daehee Park This is my paper · ORCID

classification cs.ROcs.CV

keywords motionpredictionplanningmodelmergingskillconflictdisjointparametertrainingsparsecrowdnavigationedgedeployment

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Unified robot navigation models that share one small encoder for motion prediction and motion planning suffer from what the paper calls Skill Conflict: the two tasks' gradient updates land on the same weights, so strengthening one skill degrades the other and joint performance saturates early. To fix this, the paper proposes Disjoint Parameter Training (DPT), which alternately trains each task and permanently assigns it a distinct, non-overlapping set of high-gradient parameters, then freezes those masks so each task's core skill stays local. After that, Sparse Merging adds back only the most salient 1% of each task vector to the pretrained weights, suppressing interference between adjacent features. On crowd-navigation benchmarks, this framework cuts planning average displacement error roughly in half compared with a leading unified baseline while using only about 6.4M parameters, and an optional joint-reasoning stage improves both tasks further. A sympathetic reader would take away that parameter-level separation, not model size, is the lever that makes compact unified prediction and planning work.

What carries the argument

The load-bearing objects are the task vector τ_t = Θ_t − Θ_0, the binary activation mask M_t^(K) that selects top-K coordinates by magnitude, and the alternating DPT loop that grows disjoint masks from unassigned parameters. The merge formula Θ_merged = Θ_0 + M_pred^(K%) ⊙ τ_pred + M_plan^(K%) ⊙ τ_plan is the single operation that does the work: disjoint masks guarantee spatial separation of core skills, while the 1% sparsity filters out low-magnitude adjacent-feature coupling that would otherwise cause cross-task interference. The paper also uses the Overlap(K) metric (coordinate-wise AND of the two masks) as the diagnostic and the controlled-overlap experiment as the causal test.

What would settle it

Replace the top-K gradient selection in Algorithm 1 with random selection of the same mask sizes on JRDB; if the random-mask DPT still delivers the reported ADE/collision gains, then magnitude saliency is not the cause — and if the controlled-overlap Table 5 trend disappears under random masks, Skill Conflict is not the mechanism.

Watch

Extended reading notes

Core claim

The paper's central discovery is a mechanism, not just a method: when prediction and planning are trained in a compact shared encoder, the top-magnitude parameter regions that each task needs overlap heavily (8% overlap at 10% utilization, 39% at 50%), and this overlap is causally tied to degraded joint performance. DPT prevents the overlap by construction: during the first half of fine-tuning, the algorithm alternately selects the top-K unassigned parameters by gradient magnitude, assigns them exclusively to one task via a binary mask, and updates only those; in the second half the masks freeze and each task consolidates. Sparse Merging then combines the two models as Θ_merged = Θ_0 + M_pre

Load-bearing premise

The load-bearing premise is that a task's 'core skill' is captured by the parameters with the largest gradient magnitudes during alternating training, so that freezing 99% of the model in sparse merging discards only irrelevant weights; if a task's information is broadly distributed or lives in low-magnitude co-adapting coordinates, the disjoint-top-K recipe would throw away needed capability.

Editorial extensions

If this is right

DPT is orthogonal to the merging rule: adding it to Task Arithmetic, Ties Merging, and T-Switch consistently improves both planning and prediction on JRDB and JTA, so merging methods designed for large models can be rescued for compact models by changing how material models are trained.
Sparse merging is not a detail: K=1–2% consistently beats dense merging (K=100%) and moderate sparsity, implying that in compact encoders each task's transferable skill lives in a narrow coordinate set.
Because the merged model is a single network with no added inference branches, it keeps edge compatibility: roughly 6.4M parameters, 10.4 ms latency, 24.6 MB model size, and 1.96 MB inference memory on an RTX 3090.
An optional joint-reasoning fine-tuning stage on the unassigned parameter regions recovers the benefit of unified reasoning (ADE 0.3640, collision 0.0074) without sacrificing the disjoint skill separation.
The same recipe transfers beyond social navigation: applying DPT and sparse merging to an end-to-end autonomous driving model improved closed-loop and open-loop planning/perception metrics on Bench2Drive.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the magnitude-saliency premise is right, a testable extension follows: DPT should combine cleanly with post-training compression — pruning or quantizing the merged model's low-magnitude coordinates should cost little, because the skill was already concentrated in the top 1%.
The overlap diagnostic could be used as a cheap predictor before committing to DPT: compute the two top-K masks from a standard fine-tuned pair; if Overlap(K) is high, switching to disjoint training should help, and the expected gain should scale with the measured overlap.
The paper's framing suggests Skill Conflict is a general resource-constrained multi-task phenomenon, not just prediction-vs-planning: their own three-way split of prediction shows conflicts within a single task; one could extend DPT to more tasks, with the caveat that the 'sweet spot' tradeoff between specialization and capacity-sharing will need a principled allocation rule.
Because the merge formula is a simple arithmetic identity on masks and task vectors, it is architecture-agnostic; one could transfer it to other compact multi-task models (e.g., embedded vision-language-action policies) provided the top-K gradient premise holds there.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Two things to know. The headline number is real: DPT plus Sparse Merging gets a 6.4M-parameter unified model that beats much heavier baselines on JRDB planning (ADE 0.404 vs 0.805 for DIPP; collision 0.009 vs 0.017). The second is that the causal story—overlapping top-K gradient coordinates cause the degradation, and preserving them fixes it—is plausible but only partially supported.

What's new: the DPT alternating disjoint-mask fine-tuning procedure and sparse top-1% task-vector merging, applied to compact prediction/planning models. I haven't seen this combination in the model-merging or multi-task robotics literature. The empirical documentation is unusually thorough: overlap-versus-mask-ratio curves, forced-overlap causal experiment in Supp Tab 5, replication on DTPP, sensitivity sweeps over schedules and seeds, efficiency numbers, and a HiP-AD autonomous-driving extension. That's more than most papers in this area ship.

Soft spots, in proportion. The stress-test concern is legitimate: no random-mask control. All experiments select coordinates by gradient magnitude, including the causal overlap experiment, so we learn that overlap among these coordinates hurts, but not that these coordinates are task-critical. The gains could come from disjointness, from regularization toward pretrained weights, or from the DPT schedule and learning rates. That's an interpretation gap, not a fundamental flaw. There are also smaller issues: no released code, single-run main tables, and hyperparameters like K=1% and the 50:50 allocation ratio appear to be selected on the JRDB test set. The full-JTA numbers and HiP-AD results are less dramatic than the headline JRDB numbers, which I would have liked to see discussed.

For whom: people working on compact unified models for edge robotics and those applying model merging to small backbones. It deserves a serious referee. My recommendation: send to peer review, ask for code/configs, validation-based hyperparameter selection, error bars on baselines, and a random-mask control. The empirical method likely stands even if the causal story needs revising.

Referee Report

4 major / 5 minor

Summary. The paper addresses the problem of jointly performing motion prediction and motion planning in compact, shared-encoder models for social robot navigation. It identifies a phenomenon called "Skill Conflict": when the top-magnitude parameter regions used by the two tasks overlap, joint performance degrades. It provides empirical evidence for this via mask-ratio sweeps (Fig. 2), parameter-overlap visualizations (Fig. 3), and a forced-overlap experiment (Supp. Tab. 5). As a remedy, the paper proposes Disjoint Parameter Training (DPT), which trains each task on disjoint, progressively grown top-gradient-magnitude masks (Algorithm 1), followed by Sparse Merging, which keeps only the top 1% of each task vector by magnitude (Eq. 10). The paper claims large improvements over existing unified models and merging baselines on JRDB and JTA (e.g., planning ADE 0.8048→0.4044 vs. DIPP, collision rate 0.0167→0.0091), with a further joint-reasoning stage improving to 0.3640 ADE and 0.0074 collision rate. The central claim is that the gains come from preserving each task's "core skill" in a narrow set of high-magnitude parameters while suppressing cross-task interference.

Significance. If the mechanism is valid, the paper makes a practically relevant contribution: it gives one of the first systematic studies of parameter-level interference in compact unified prediction-planning models and proposes a training/merging recipe that could improve edge-deployable robot navigation. The empirical work is unusually thorough: replicate the overlap phenomenon on DTPP, provide seed and schedule sensitivity analyses, extend DPT to HiP-AD/E2E-AD, and report inference cost (10.4 ms, 6.44M params), which supports the resource-constrained motivation. The main weakness is that the causal interpretation is not yet fully isolated from the magnitude-saliency proxy shared by the diagnostic, the causal experiment, and the remedy. This is a correctness-risk concern rather than a circularity in the performance claims, because the final metrics are external and the forced-overlap experiment gives some causal grounding. The requested additional controls would determine whether the mechanism is specifically "core-skill preservation" as opposed to a more generic sparsity/regularization effect.

major comments (4)

[Sec. 3.4–3.5, Supp. Sec. 6.4 (Tab. 5)] The mechanistic interpretation is not validated independently of the magnitude-based selection. Algorithm 1 selects top-K coordinates by |gradient|, Eq. (10) keeps top-K task-vector entries by magnitude, and Supp. Tab. 5 forces overlap only among masks that were themselves chosen by gradient magnitude. This does not test whether magnitude identifies task-critical parameters. The observed gains could stem from generic sparse-mask regularization, from disjointness, or from DPT-specific hyperparameters rather than from preserving 'core' coordinates. Please add a random-mask control: run DPT with randomly selected masks of the same size and schedule, and merge random 1% subsets of task vectors. If magnitude selection is load-bearing, these controls should be clearly worse on the same metrics. This should also be applied to the diagnostic, since the overlap metric (Eq. 9) and the remedy share
[Sec. 3.5, Table 3] The text states that Sparse Merging 'consistently outperforms dense merging (K=100%) and moderately sparse configurations (e.g., K=10%)', but Table 3 stops at K=40, so the dense-merging comparison is missing. Because sparsity is the second half of the remedy, please report K=100% (and ideally intermediate values such as 60–80%) under the same DPT settings. Also, the accompanying claim that K=2 is 'best overall across all metrics' is inaccurate: in Table 3, K=1 has a lower collision rate (0.0091 vs. 0.0096), so the sentence should be rephrased to reflect the small non-monotonicity.
[Sec. 4.2, Table 1] The interpretation of the upper rows of Table 1 is unclear. The text says Plan/Pred Finetune 'performance remains largely unchanged even with DPT', but the table shows large swings in the cross-task metrics, e.g., Plan Finetune has prediction ADE 1.9825 while Plan Finetune + DPT has 0.7368, and Pred Finetune planning ADE changes from 0.8889 to 0.8469. It is not stated whether these rows are single-task fine-tuned models evaluated with both heads, merged models, or something else. Please specify exactly which parameters are trained and what is evaluated, since this affects the claim that DPT preserves each task's inherent capability and the interpretation of DPT's compatibility with other merging methods.
[Sec. 4.2 (Effect of Adding Joint Reasoning)] The joint-reasoning stage is used to obtain the best numbers in Table 4 (DPT + SM + JR), but its description is only two sentences: a fine-tuning stage with a joint task loss on 'parameter regions that remain inactive after DPT and Sparse Merging'. The loss definition, learning rate, number of epochs, whether masks are frozen, and which specific parameter regions are updated are not given either in the main text or in the supplementary implementation details. Please provide these details, since this stage is part of the headline result.

minor comments (5)

[Algorithm 1, Eq. (8)–(10)] The symbol K is overloaded: in Algorithm 1 it denotes the per-step mask size, while in Eq. (8)–(10) it denotes the Sparse Merging mask ratio. This creates confusion when both are discussed in the same section. Use distinct notation (e.g., K_dpt and K_sm).
[Figure captions] Several figure captions contain internal version/date strings that should be removed: Fig. 4 'Final Ver_251114_1853', Fig. 6 '251113 22:01', and Supp. Fig. 1 '2511191603'.
[Sec. 4.1, Table 4] The comparison against DIPP and DTPP does not state whether these baselines were re-trained under the same data splits and training protocol or taken from official checkpoints. Without this, differences in preprocessing or training budget may confound the architecture comparison.
[Eq. (9)] The quantity D is defined as the number of parameters activated by at least one task, which is itself a function of K. This notation is easy to confuse with dataset size or embedding dimension; rename it (e.g., N_union) and state the dependence on K explicitly.
[Sec. 4.2, Table 1] Minor wording issue: 'the fact that their performance remains largely unchanged even with DPT' does not match the large cross-task improvements in the table. Even if the intended meaning is 'single-task capability is preserved', the sentence should be rewritten to avoid the contradiction.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the shared top-K magnitude assumption is a validation gap, not a constructional equivalence, and headline results are externally benchmarked.

full rationale

After walking the derivation chain, I find no step where a claimed prediction or explanation reduces by construction to an input. The Skill Conflict diagnostic (Eq. 9) measures coordinate-wise overlap of top-K magnitude masks, and Sparse Merging (Eq. 10) selects the same top-K coordinates, so the diagnosis and remedy share the magnitude-saliency assumption. But sharing an assumption is not circular equivalence: the forced-overlap experiment (Supp. Sec. 6.4, Tab. 5) manipulates overlap while holding training settings fixed and observes benchmark performance changes, and all headline numbers (Tab. 1, Tab. 4) are measured on JRDB/JTA, not derived from Eq. 9 or Eq. 10. The strongest concern, that no random-mask control validates gradient magnitude as task-critical, is an internal validation gap about the saliency proxy, not a constructional equivalence; the paper's explicit scope limitations (Supp. Sec. 6.2, Sec. 10) acknowledge capacity/settings limits rather than conceal a fitted-parameter-as-prediction step. No load-bearing self-citation or imported uniqueness theorem appears: references to the authors' prior work (refs. 26, 27, 40) are contextual. Score 0.

Assumptions & free parameters 5 free parameters · 5 assumptions · 1 invented entities

The paper's central claim rests on several domain assumptions about task vectors, magnitude-based saliency, the value of game-theoretic pretraining, and the validity of offline benchmarks. The main free parameters — K, DPT allocation, schedule, and per-task learning rates — were tuned on the evaluation benchmarks, and the collision-cost constants are unreported, which raises the burden on reproduction.

free parameters (5)

Sparse Merging mask ratio K = 1% (swept 1-40%; best around 1-2%)
Tab. 3; determines how many top-magnitude task-vector coordinates survive merging, used for all headline results and in Eq. 10.
DPT allocation ratio (plan:pred) = 50:50 (swept 90:10 to 10:90)
Tab. 2 and Supp Tab. 3; controls maximum learnable parameter share per task and is performance-tradeoff sensitive.
DPT mask growth:frozen schedule = 1:1 (swept 1:3, 1:1, 3:1)
Supp Tab. 2; the first half grows masks, the second half freezes them; 1:1 is used in main experiments.
DPT learning rates per task = planner 5e-6, predictor 2e-6 (JRDB); both 1e-5 (JTA)
Supp Sec. 4; asymmetric learning rates affect which gradients dominate top-K selection and the final skill balance.
Collision cost hyperparameters λ_col, τ, ε = not reported
Supp Eqs. 1-10; these shape the game-theoretic pretraining and task fine-tuning losses; without their values the training objective is not fully specified.

assumptions (5)

domain assumption Task vectors τ_t = Θ_t - Θ_0 additively encode task-specific skills in compact transformers.
Sec. 3.2 Eq. 6; the entire merging pipeline assumes task-vector directions can be masked and linearly recombined.
ad hoc to paper Top-K gradient/task-vector magnitude selects the task-critical parameters.
Algorithm 1 line 6 and Eqs. 8/10; both DPT and Sparse Merging assume magnitude-based saliency, supported only by internal causal experiments.
domain assumption A single game-theoretically pretrained Θ0 retains both task capabilities and serves as a common initialization.
Sec. 3.4 and Supp Sec. 2; if Θ0 already solves both tasks, DPT mostly preserves pretrained skills rather than learning them from scratch; no ablation with a different pretraining target is reported.
ad hoc to paper Sparse merging with K≈1-2% preserves each task's core skill while suppressing interference.
Eq. 10 and Tab. 3; tested only after DPT and on JRDB sweeps, then used for JTA and E2E-AD without separate validation.
domain assumption Offline JRDB/JTA metrics measured against ground-truth future trajectories are valid proxies for safe crowd navigation.
Sec. 4.1 metrics; no closed-loop crowd evaluation is reported, and collision rate uses ground-truth future paths, not predicted agents.

invented entities (1)

Skill Conflict
purpose: Conceptual label and quantitative target for inter-task parameter overlap in compact unified encoders (Eq. 9), used to motivate DPT and Sparse Merging.
Defined in terms of top-K task-vector masks chosen by the authors; the causal experiment (Supp Sec. 6.4) is internal to the framework, so there is no external falsifiable handle beyond this paper's own benchmarks.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Unified Prediction and Planning via Conflict-Aware Disjoint Parameter Training." pith.science (2026). https://pith.science/paper/P56HPZOS

@misc{pith2026260719971,
  author       = {Pith},
  title        = {Pith review of: Unified Prediction and Planning via Conflict-Aware Disjoint Parameter Training},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/P56HPZOS}},
  note         = {Machine review of arXiv:2607.19971}
}

abstract

Accurate motion prediction of surrounding agents and safe motion planning are two closely coupled key tasks for social robot navigation in crowded environments. Deploying these systems on resource-constrained edge devices necessitates compact, unified models that can perform both tasks simultaneously. However, within these compact shared encoders, recent unified models often overlook severe representational conflicts that arise from the distinct objectives of predicting neighbor behaviors versus ego-centric safety planning. To address this issue, we first identify the Skill Conflict$\unicode{x2014}$a phenomenon where overlapping parameter assignments cause distinct tasks to compete for the same weights, preventing the model from fully specializing in individual skills. To resolve this, we propose a novel model-merging-based framework, Disjoint Parameter Training (DPT). DPT mitigates performance degradation caused by Skill Conflict through distributed parameter learning, which separates the key parameter regions of each task while preserving their core capabilities prior to merging. In addition, we observe that sparse merging, which selectively integrates only the most influential parameters for each task rather than combining all task-specific parameters, yields optimal performance by preventing interference among adjacent features and concentrating representational capacity. DPT can be applied in parallel with a variety of merging methods. Evaluated on standard crowd navigation benchmarks (JRDB and JTA), our framework demonstrates superior performance, validating its versatility and effectiveness for safe, resource-efficient robot navigation.

Figures

Figures reproduced from arXiv: 2607.19971 by the authors.

**Figure 1.** Comparison of learning strategies in unified models. (a) In previous methods, the skills of each task are learned within a single shared encoder, causing conflicts between the two skills in the parameter space. (b) DPT resolves this issue by separately training the major parameter regions allocated to each task and Sparse Merging combines only their core parameter regions, thereby mitigating skill conflicts within… view at source ↗

**Figure 2.** Planning and Prediction performance according to the mask ratio K. Plan Merged and Pred Merged use the single task vectors corresponding to the mask ratio. In Both Merged, the parameter utilization of the non-target task is fixed at 40%, while that of the target task is used according to the given parameter ratio [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 4.** Overview of our proposed framework. In the DPT step, task-specific models for prediction and planning are generated by separating training regions with binary masks to minimize skill conflict. In the Sparse Merging, the two fine-tuned models are combined into a single unified model by utilizing only the most significant task vectors, effectively preserving each task’s skill. 3.4 DPT: Preparing Task-Specialized Mater… view at source ↗

Figures from the paper (10 more)

**Figure 5.** Figure 5: Performance of (a) Planning and (b) Prediction tasks trained with DPT depends on mask ratio K. Task-Specific Parameter Allocation of DPT. DPT performance is subject to variation based on the extent of task-specific parameter allocation. As shown in Tab. 2, we conducte…

**Figure 6.** Figure 6: Qualitative Results on the JRDB dataset. (a)–(b) show unified models with different architectures, while (c)–(e) present results of applying various baseline model merging methods on our backbone-based unified model. These are compared with (f) Ours, which employs DPT …

**Figure 1.** Figure 1: Overview of the unified backbone architecture. Our model extends the SocialTransmotion backbone by preserving its multi-modal tokenization—trajectory, pose, and bounding box embeddings—and processing all agent-wise tokens through a shared Transformer encoder. There ar…

**Figure 2.** Figure 2 [PITH_FULL_IMAGE:figures/full_fig_p027_2.png]

**Figure 3.** Figure 3: Complementary Qualitative Results on the JRDB [PITH_FULL_IMAGE:figures/full_fig_p027_3.png]

**Figure 4.** Figure 4: (a) Parameter overlap ratio between planning and prediction tasks. (b)-(c) Planning and prediction performance for single-task merging and joint-task merging on the DTPP architecture [PITH_FULL_IMAGE:figures/full_fig_p028_4.png]

**Figure 5.** Figure 5: Parameter distribution by layer group and the degree of skill conflict between Plan Skill and Pred Skill in the DTPP architecture at mask ratios of K=10% and K=50%. 6 Empirical Study 6.1 Skill Conflict in Other Architecture To verify that the Skill Conflict analysis co…

**Figure 6.** Figure 6: Effect of model dimension on skill conflict. The overlap ratio is measured under a fixed mask ratio of K=50%. Compact models with dimensions from 32 to 128 exhibit consistently high overlap ratios of 41–42%, while larger models with dimensions of 256 and 512 show reduc…

**Figure 7.** Figure 7: Evolution of disjoint masks across DPT allocation steps [PITH_FULL_IMAGE:figures/full_fig_p032_7.png]

**Figure 8.** Figure 8: Layer-wise distribution of Skill Conflict [PITH_FULL_IMAGE:figures/full_fig_p033_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

17 extracted references

[1]

In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

Hu, Y., Yang, J., Chen, L., Li, K., Sima, C., Zhu, X., Chai, S., Du, S., Lin, T., Wang, W., et al.: Planning-oriented autonomous driving. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 17853– 17862 (2023)

2023
[2]

In: 2024 IEEE International Conference on Robotics and Automation (ICRA)

Huang, Z., Karkus, P., Ivanovic, B., Chen, Y., Pavone, M., Lv, C.: DTPP: Differ- entiable joint conditional prediction and cost evaluation for tree policy planning in autonomous driving. In: 2024 IEEE International Conference on Robotics and Automation (ICRA). pp. 6806–6812. IEEE (2024)

2024
[3]

In: Proceedings of the IEEE/CVF International Conference on Computer Vision

Huang, Z., Liu, H., Lv, C.: GameFormer: Game-theoretic modeling and learning of transformer-based interactive prediction and planning for autonomous driving. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 3903–3913 (2023)

2023
[4]

Advances in Neural Information Processing Systems37, 819–844 (2024)

Jia, X., Yang, Z., Li, Q., Zhang, Z., Yan, J.: Bench2Drive: Towards multi-ability benchmarking of closed-loop end-to-end autonomous driving. Advances in Neural Information Processing Systems37, 819–844 (2024)

2024
[5]

In: The Thirteenth International Conference on Learning Representations (2025)

Jia, X., You, J., Zhang, Z., Yan, J.: DriveTransformer: Unified transformer for scal- able end-to-end autonomous driving. In: The Thirteenth International Conference on Learning Representations (2025)

2025
[6]

In: Proceedings of the IEEE/CVF International Conference on Computer Vi- sion

Jiang,B.,Chen,S.,Xu,Q.,Liao,B.,Chen,J.,Zhou,H.,Zhang,Q.,Liu,W.,Huang, C., Wang, X.: VAD: Vectorized scene representation for efficient autonomous driv- ing. In: Proceedings of the IEEE/CVF International Conference on Computer Vi- sion. pp. 8340–8350 (2023)

2023
[7]

In: 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

Kedia, K., Dan, P., Choudhury, S.: A game-theoretic framework for joint fore- casting and planning. In: 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). pp. 6773–6778. IEEE (2023)

2023
[8]

In: Proceedings of the IEEE/CVF International Conference on Computer Vision

Lee, G., Jeong, W., Park, D., Jeong, J., Yoon, K.J.: Interaction-merged motion planning: Effectively leveraging diverse motion datasets for robust planning. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 28610–28621 (2025)

2025

Show all 17 references

[9]

In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

Liao, B., Chen, S., Yin, H., Jiang, B., Wang, C., Yan, S., Zhang, X., Li, X., Zhang, Y., Zhang, Q., et al.: DiffusionDrive: Truncated diffusion model for end-to-end autonomous driving. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 1...

2025
[10]

Advances in Neural Information Processing Systems34, 18878–18890 (2021)

Liu, B., Liu, X., Jin, X., Stone, P., Liu, Q.: Conflict-averse gradient descent for multi-task learning. Advances in Neural Information Processing Systems34, 18878–18890 (2021)

2021
[11]

In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

Qi, B., Li, F., Wang, Z., Gao, J., Li, D., Ye, P., Zhou, B.: Less is more: Effi- cient model merging with binary task switch. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 15265–15274 (2025) 38 Seo et al

2025
[12]

In: The Twelfth International Conference on Learning Representations (2024)

Saadatnejad, S., Gao, Y., Messaoud, K., Alahi, A.: Social-Transmotion: Prompt- able human trajectory prediction. In: The Twelfth International Conference on Learning Representations (2024)

2024
[13]

In: 2025 IEEE International Conference on Robotics and Automation (ICRA)

Sun, W., Lin, X., Shi, Y., Zhang, C., Wu, H., Zheng, S.: SparseDrive: End-to-end autonomous driving via sparse scene representation. In: 2025 IEEE International Conference on Robotics and Automation (ICRA). pp. 8795–8801. IEEE (2025)

2025
[14]

In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)

Tang, Y., Xu, Z., Meng, Z., Cheng, E.: HiP-AD: Hierarchical and multi-granularity planning with deformable attention for autonomous driving in a single decoder. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). pp. 25605–25615 (October 2025)

2025
[15]

In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

Weng, X., Ivanovic, B., Wang, Y., Wang, Y., Pavone, M.: PARA-Drive: Parallelized architecture for real-time autonomous driving. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 15449–15458 (2024)

2024
[16]

Advances in Neural Information Processing Systems36, 7093–7115 (2023)

Yadav, P., Tam, D., Choshen, L., Raffel, C.A., Bansal, M.: TIES-Merging: Resolv- ing interference when merging models. Advances in Neural Information Processing Systems36, 7093–7115 (2023)

2023
[17]

Advances in Neural Information Processing Systems33, 5824–5836 (2020)

Yu, T., Kumar, S., Gupta, A., Levine, S., Hausman, K., Finn, C.: Gradient surgery for multi-task learning. Advances in Neural Information Processing Systems33, 5824–5836 (2020)

2020

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.

[1] [1]

In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

Hu, Y., Yang, J., Chen, L., Li, K., Sima, C., Zhu, X., Chai, S., Du, S., Lin, T., Wang, W., et al.: Planning-oriented autonomous driving. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 17853– 17862 (2023)

2023

[2] [2]

In: 2024 IEEE International Conference on Robotics and Automation (ICRA)

Huang, Z., Karkus, P., Ivanovic, B., Chen, Y., Pavone, M., Lv, C.: DTPP: Differ- entiable joint conditional prediction and cost evaluation for tree policy planning in autonomous driving. In: 2024 IEEE International Conference on Robotics and Automation (ICRA). pp. 6806–6812. IEEE (2024)

2024

[3] [3]

In: Proceedings of the IEEE/CVF International Conference on Computer Vision

Huang, Z., Liu, H., Lv, C.: GameFormer: Game-theoretic modeling and learning of transformer-based interactive prediction and planning for autonomous driving. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 3903–3913 (2023)

2023

[4] [4]

Advances in Neural Information Processing Systems37, 819–844 (2024)

Jia, X., Yang, Z., Li, Q., Zhang, Z., Yan, J.: Bench2Drive: Towards multi-ability benchmarking of closed-loop end-to-end autonomous driving. Advances in Neural Information Processing Systems37, 819–844 (2024)

2024

[5] [5]

In: The Thirteenth International Conference on Learning Representations (2025)

Jia, X., You, J., Zhang, Z., Yan, J.: DriveTransformer: Unified transformer for scal- able end-to-end autonomous driving. In: The Thirteenth International Conference on Learning Representations (2025)

2025

[6] [6]

In: Proceedings of the IEEE/CVF International Conference on Computer Vi- sion

Jiang,B.,Chen,S.,Xu,Q.,Liao,B.,Chen,J.,Zhou,H.,Zhang,Q.,Liu,W.,Huang, C., Wang, X.: VAD: Vectorized scene representation for efficient autonomous driv- ing. In: Proceedings of the IEEE/CVF International Conference on Computer Vi- sion. pp. 8340–8350 (2023)

2023

[7] [7]

In: 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

Kedia, K., Dan, P., Choudhury, S.: A game-theoretic framework for joint fore- casting and planning. In: 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). pp. 6773–6778. IEEE (2023)

2023

[8] [8]

In: Proceedings of the IEEE/CVF International Conference on Computer Vision

Lee, G., Jeong, W., Park, D., Jeong, J., Yoon, K.J.: Interaction-merged motion planning: Effectively leveraging diverse motion datasets for robust planning. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 28610–28621 (2025)

2025