YOUDrive: Driving Style as a Composable Direction for End-to-End Autonomous Driving

Guantian Zheng; Lidong Yu; Tianyu Gao; Jiaxing Chen; Jiashu Li; Chen Lv

cs.CVOct 10, 2026

YOUDrive: Driving Style as a Composable Direction for End-to-End Autonomous Driving

Authors: Guantian Zheng, Lidong Yu, Tianyu Gao, Jiaxing Chen, Jiashu Li, Chen Lv

Organizations: AutoMan Lab, Nanyang Technological University · Neolix · City University of Hong Kong · School of Vehicle and Mobility, Tsinghua University

Abstract

End-to-end driving policies can achieve strong task performance while offering limited control over their driving style. Driving style describes how a vehicle accelerates, brakes, and maintains distance. Policies trained without explicit preference conditioning lack a direct mechanism for adjusting these behaviors at inference. We propose YOUDrive, a vision--language--action planner for continuous and composable driving-style control. Each teacher's behavior is distilled into a task vector, a learned parameter update that induces a style direction in the planner's representation. These directions can be scaled and combined at inference without retraining the backbone. Conditioned on these directions, a flow-matching decoder models a multimodal trajectory distribution and generates kinematically feasible motion alternatives. We introduce the Style Alignment Score (SAS) to describe human-referenced kinematic behavior after safety and task gating. On NAVSIM navtest, a single coefficient monotonically controls assertiveness across six teacher-derived personas, while PDMS remains above 0.90 across the tested individual and blended configurations, reaching 0.954. Further analyses reveal systematic policy-specific kinematics after controlling for scene context, whereas the evaluated discrete style labels provide limited additional explanatory power. Code, checkpoints, and the metric will be released.

Figures & tables

Figure 1: YOUDrive pipeline. The VLA backbone provides content tokens, style is a task-vector interpolation applied to those tokens, and a flow-matching decoder samples several feasible candidates before returning one.

each teacher’s native style ( $\alpha{=}$ T) and our persona steered to it ( $\alpha{=}1$ )
			PDMS sub-scores					SAS
config	$\alpha$	PDMS $\uparrow$	NC	DAC	EP	TTC	Comf	$H$
human anchor	–	–	–	–	–	–	–	0.00
base (ours)	0	0.912	98.8	96.6	86.7	96.3	100.0	0.48
GTRS	T	0.899	99.3	97.6	83.2	96.1	100.0	$-0.01$
$\hookrightarrow$ ours	1.0	0.903	97.5	96.6	87.6	93.6	100.0	0.05

Table 1: PDMS with its sub-scores (no-collision NC, drivable-area compliance DAC, ego-progress EP, time-to-collision TTC, comfort Comf) and the SAS assertiveness axis

H

( Eq. 9 ) on NAVSIM navtest. For each teacher we list its native behavior (

\alpha{=}

T) and, below it (

\hookrightarrow

ours), our persona steered to that teacher at full strength (

\alpha{=}1

); base is the same model with no persona (

\alpha{=}0

), and the human driver sits at

H{=}0

. Native-teacher PDMS are the reported navtest v1 numbers from each planner’s paper [ 3 , 17 , 25 , 19 , 16 ] ; GTRS reports v2 EPDMS only, so its v1 row is our own evaluation.

kinematic feature	$\eta^{2}$ (driver) $\uparrow$	ICC(1)	$F$
avg. speed	0.005	0.005	20
peak accel	0.275	0.288	1631
peak decel	0.287	0.300	1735
long. jerk	0.445	0.461	3451
lat. jerk	0.460	0.476	3663
lat. accel (max)	0.167	0.176	862

Table 2: Variance decomposition of per-token kinematics with driver identity as the factor (16 drivers, common tokens). Control features are driver-dominated, while average speed is scene-locked. In contrast, the coarse ANC label of [ 6 ] carries little style signal once environment is accounted for ( Sec. 4.4 ).

Figure 2: Style geometry in atom space (human-referenced signed $z$ ; the dotted ring marks the human reference, radius grows with assertiveness on the dynamics half $H$ and with gap-tightening on the proximity half $P$ ). Left: native teachers ( $\alpha{=}$ T); right: our personas. Both fan out along the dynamics atoms (the primary axis $H$ ); on the proximity atoms the cross-policy spread is small, so $P$ is near-degenerate on navtest and secondary, with a teacher/persona sign mismatch we treat as an open problem ( Sec. 4.7 ).

	ANC label	driver identity
raw effect size, peak accel	0.006	0.275
raw effect size, peak decel	0.012	0.287
raw effect size, assertiveness $H$	0.063	0.270
$\Delta R^{2}(\cdot\mid\text{env})$ on $H$	0.037	0.232
$\Delta R^{2}(\cdot\mid\text{env})$ on $P$	0.000	0.065

Table 3: Two complementary diagnostics of what organizes driving style, on the navtest atoms. The ANC column groups the human log-replay trajectories by StyleDrive’s label; the driver column groups the 16 policies’ trajectories by identity. Top: raw between-group effect size (

\varepsilon^{2}

, rank-based, for ANC;

\eta^{2}

, one-way ANOVA, for driver). Bottom: incremental adjusted

R^{2}

after regressing out the scene environment, penalizing the differing degrees of freedom (3 ANC classes vs. 16 drivers). After control, model identity remains a strong organizer, unlike the label;

P

is weak for both ( Sec. 4.7 ).

Figure 3: Style controllability in BEV. Each row shows one scene decoded at increasing style strength $\alpha$ (columns, $0\rightarrow 1$ ): an assertive persona (DiffusionDriveV2, top) and a cautious persona (GTRS, bottom). Trajectory color encodes $\alpha$ ( $0$ is the base output, $1$ is full persona style); the black curve is the fixed base anchor. A single coefficient sweeps the decoded trajectory continuously, and every $\alpha$ stays safe (gate-passing). Ego is outlined in red, other agents in blue.

Figure 4: Dose–response: as the single strength $\alpha$ rises, the dynamics atoms (peak deceleration and longitudinal jerk shown) move monotonically from the base toward each teacher’s operating point, so $\alpha$ is a continuous style coefficient rather than a discrete selector. The smoother personas (lower panel, zoomed) separate cleanly at small magnitudes.

Table 4: Style following as

\alpha

increases (full sweep in Tab. 7 ). One coefficient

\alpha

steers the base toward each teacher in opposite directions on the assertiveness axis

H

—the aggressive DiffusionDriveV2 persona (left) rises while the smooth GTRS persona (right) falls—with PDMS high throughout.

Table 5: Persona directions behave as reusable, composable task vectors.

Table 6: Ablations: decoder family (a) and inference budget (b). Sub-scores as in Tab. 1 .

Appendix figures & tables2 assets

Supplementary material from the paper’s appendix.

Appendix

scope	teacher	$\alpha$	PDMS $\uparrow$	$H$	$P$	$\bar{v}$	$a^{+}$	$a^{-}$	$j_{\ell}$	$j_{t}$	thw	ttc	lead	$d_{\min}$	$d_{\mathrm{vru}}$
anchor	human	–	0.9480	+0.00	+0.00	4.6181	0.2446	0.3059	0.2898	0.1026	3.7850	4.3915	12.9060	4.5340	12.0630
anchor	base(ours)	–	0.9120	+0.48	+0.08	4.6051	0.5282	0.6151	0.8865	0.4750	3.7640	4.3710	12.8275	4.5030	12.0780
sweep	gtrs	0	0.9120	+0.48	+0.08	4.6051	0.5282	0.6151	0.8865	0.4750	3.7640	4.3710	12.8275	4.5030	12.0780
sweep	gtrs	0.25	0.9187	+0.27	+1.75	4.8300	0.3713	0.3331	0.4590	0.5258	3.2140	3.5045	11.6580	4.4240	11.9630
sweep	gtrs	0.5	0.9230	+0.18	+2.03	4.9096	0.3341	0.2454	0.3701	0.4722	3.1210	3.3810	11.5660	4.3940	11.9520
sweep	gtrs	0.75	0.9158	+0.10	+2.38	4.9346	0.2753	0.1638	0.3118	0.4139	3.0620	3.2390	11.1695	4.3610	11.8570

Appendix

Table 7: Full v1 navtest results for per-teacher

\alpha

sweeps, aggregated across eight shards with weights proportional to the number of scenarios. The human serves as the SAS reference (

H=P=0

), while the base is the shared policy at

\alpha=0

;

\alpha=\mathrm{T}

denotes the native teacher.

H

and

P

are the SAS axes defined in Sec. 4.2 . Sub-metrics:

\bar{v}

mean speed;

a^{+}/a^{-}

peak acceleration/deceleration;

j_{\ell}/j_{t}

longitudinal/lateral jerk; thw/ttc minimum headway/time-to-collision; lead/

d_{\min}/d_{\mathrm{vru}}

minimum lead/object/VRU distances.

scope	teacher	$\alpha$	PDMS $\uparrow$	$H$	$P$	$\bar{v}$	$a^{+}$	$a^{-}$	$j_{\ell}$	$j_{t}$	thw	ttc	lead	$d_{\min}$	$d_{\mathrm{vru}}$
combo	ddv2+goalflow 50/50	0	0.9120	+0.48	+0.08	4.6051	0.5282	0.6151	0.8865	0.4750	3.7640	4.3710	12.8275	4.5030	12.0780
combo	ddv2+goalflow 50/50	0.5	0.9520	+1.06	+1.73	4.8950	1.0565	0.9788	1.7424	0.7410	3.3390	3.7605	12.0920	4.4030	11.6645
combo	ddv2+goalflow 50/50	1	0.9500	+2.64	+2.45	4.8309	2.4876	1.8331	3.6637	1.7195	3.0040	3.4865	12.5640	4.3570	11.6665
combo	ddv2+transfuser 70/30	0	0.9120	+0.48	+0.08	4.6051	0.5282	0.6151	0.8865	0.4750	3.7640	4.3710	12.8275	4.5030	12.0780
combo	ddv2+transfuser 70/30	0.5	0.9437	+1.06	+1.76	4.7260	1.0065	0.9937	1.6054	0.8264	3.3110	3.9040	12.8625	4.2290	11.8385
combo	ddv2+transfuser 70/30	1	0.9480	+2.92	+2.76	4.8494	2.5248	2.0674	3.7133	2.1054	3.1130	3.3615	12.1140	4.1460	11.7070

Appendix

Table 8: Full v1 navtest results for persona blends. Each configuration specifies the teacher task vectors and their mixing weights, with

\alpha

controlling the overall strength. We report PDMS, SAS axes

H

and

P

, and the individual kinematic and proximity measurements, following the same definitions as Tab. 7 . All configurations recover the shared base at

\alpha{=}0

Autonomous Driving

Autonomous driving develops vehicles that perceive road scenes, predict interactions, and plan and control safe movement with limited or no human input.

See all 536 papers

STAGE: STyle-controllable Action GEneration for personalized autonomous driving
Jul 31, 2026Zihao Liu, Xing Liu, Yizhai Zhang +1Autonomous Driving
Research Center for Intelligent Robotics, School of Astronautics, Northwestern Polytechnical University, and National Key Laboratory of Aerospace Flight Dynamics, Northwestern Polytechnical University, Xi’an, China
D $^3$ -MoE:Dual Disentangled Diffusion Mixture-of-Experts for Style-Controllable End-to-End Autonomous Driving
Jun 3, 2026Renju Feng, Rukang Wang, Ning Xi +4End-to-End Autonomous Driving Diffusion-Based Planning
Intelligent Transportation Systems Research Center, Wuhan University of Technology, Wuhan, Hubei, China · School of Mechanical and Electronic Engineering, Wuhan University of Technology, Wuhan, Hubei, China · School of Computer Science and Artificial Intelligence, Wuhan University of Technology, Wuhan, Hubei, China +1
StyleVLA: Driving Style-Aware Vision Language Action Model for Autonomous Driving
Mar 10, 2026Yuan Gao, Dengyuan Hua, Mattia Piccinini +4Autonomous Driving Vision-Language-Action Models
Y. Gao, D. Hua, M. Piccinini, F. Schäfer, K. Moller, J. Betz are with the Professorship of Autonomous Vehicle Systems, TUM School of Engineering and Design, Technical University of Munich, 85748 Garching, Germany; Munich Institute of Robotics and Machine Intelligence (MIRMI) · L. Li is with the School of Mechanical and Aerospace Engineering, Nanyang Technological University

Jul 31, 2026J/K move · Enter open · S save

STAGE: STyle-controllable Action GEneration for personalized autonomous driving

Zihao Liu, Xing Liu, Yizhai Zhang, Panfeng Huang

Research Center for Intelligent Robotics, School of Astronautics, Northwestern Polytechnical University, and National Key Laboratory of Aerospace Flight Dynamics, Northwestern Polytechnical University, Xi’an, China

Autonomous Driving

Open paper AskPDF arXiv

Driving style refers to the behavioral preferences that drivers maintain during driving, shaped by their diverse experiences, habits, and needs, and is typically reflected in varying levels of aggressiveness. If humans choose to use autonomous driving systems, they would expect the driving style of the systems to closely resemble their own habit. However, this is challenging for current industrial autonomous driving systems. To address this, we developed a style controllable action generation method, STAGE, for driving tasks. Its training process is based on imitation learning, incorporating both style value and latent value action modality encoding. Preference learning is then used to identify the user's driving style as a continuous, monotonic style value. And to reduce the cost of human involvement in the preference training process, we also developed a set of rules to compare driving style in data pairs. Then, during inference, the user inputs the style value to control the generated action patterns, dynamically meeting the user's expectations. Using the STAGE method, we verified that the style-controlled action generation results in several typical road scenarios significantly align with human expectations. Furthermore, through comparisons between the STAGE method and various other approaches, we reveal the unique functionalities of STAGE, including its style controllability, style continuity, driving style alignment capability and driving safety. The code for this work is available at: https://github.com/CarlDegio/STAGE

Abstract

Autonomous Driving

Explore similar work

STAGE: STyle-controllable Action GEneration for personalized autonomous driving

D3^33-MoE:Dual Disentangled Diffusion Mixture-of-Experts for Style-Controllable End-to-End Autonomous Driving

StyleVLA: Driving Style-Aware Vision Language Action Model for Autonomous Driving

D $^3$ -MoE:Dual Disentangled Diffusion Mixture-of-Experts for Style-Controllable End-to-End Autonomous Driving