{"total":11,"items":[{"citing_arxiv_id":"2605.19771","ref_index":47,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Beyond Imitation: Learning Safe End-to-End Autonomous Driving from Hard Negatives","primary_cat":"cs.RO","submitted_at":"2026-05-19T12:41:47+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"BeyondDrive augments imitation learning with synthesized safety-critical negative trajectories and a repulsive loss to improve safety in autonomous driving, reporting 89.7 PDMS on NAVSIMv1 and generalization to other models.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.15120","ref_index":38,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"CLOVER: Closed-Loop Value Estimation and Ranking for End-to-End Autonomous Driving Planning","primary_cat":"cs.RO","submitted_at":"2026-05-14T17:32:18+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"CLOVER is a closed-loop generator-scorer framework that expands proposal coverage with pseudo-expert trajectories and performs conservative self-distillation to achieve state-of-the-art planning scores on NAVSIM and nuScenes.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.09701","ref_index":55,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"DriveFuture: Future-Aware Latent World Models for Autonomous Driving","primary_cat":"cs.CV","submitted_at":"2026-05-10T18:45:21+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"DriveFuture achieves SOTA results on NAVSIM by conditioning latent world model states on future predictions to directly inform trajectory planning.","context_count":1,"top_context_role":"baseline","top_context_polarity":"baseline","context_text":"3 94.9 91.5 97.7 67.8 27.1Stage 2 87.3 76.7 88.8 99.2 84.3 85.1 49.7 93.1 44.5 Senna-E2E [53] ResNet-50 Stage 1 95.6 86.0 98.9 99.6 83.9 95.1 95.3 97.6 75.6 27.2Stage 2 78.6 74.8 84.8 98.2 88.2 75.7 46.9 96.0 65.8 DriveSuprim [54] V2-99 Stage 1 98.9 95.1 99.2 99.6 76.1 99.1 94.7 97.6 54.2 42.1Stage 2 87.9 88.8 89.6 98.8 80.3 86.0 53.5 97.1 56.1 ZTRS [55] V2-99 Stage 1 98.9 97.6 100.0 100.0 66.7 98.9 96.2 96.7 44.0 48.1Stage 2 91.1 90.4 95.8 99.0 63.6 89.8 60.4 97.6 66.1 GTRS-E [49] V2-99+EV A-ViT-L+ViT-LStage 1 98.9 99.3 99.8 99.8 75.2 98.4 96.0 97.6 51.6 49.4Stage 2 92.3 93.3 94.6 99.2 73.1 91.2 53.9 96.7 56.8 SimScale [56] V2-99 Stage 1 99.6 99.1 99.9 100.0 69.6 99.6 95.8 95.6 28.4 53.2Stage 2 94."},{"citing_arxiv_id":"2605.00066","ref_index":21,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Do Open-Loop Metrics Predict Closed-Loop Driving? A Cross-Benchmark Correlation Study of NAVSIM and Bench2Drive","primary_cat":"cs.RO","submitted_at":"2026-04-30T09:27:57+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":4.0,"formal_verification":"none","one_line_summary":"Cross-benchmark analysis of 8 methods shows NAVSIM PDM Score correlates with Bench2Drive Driving Score at Spearman ρ=0.90, with Ego Progress as the strongest single predictor and a simpler 3-metric formula matching the full score.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.19710","ref_index":75,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"SpanVLA: Efficient Action Bridging and Learning from Negative-Recovery Samples for Vision-Language-Action Model","primary_cat":"cs.CV","submitted_at":"2026-04-21T17:34:19+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"SpanVLA reduces action generation latency via flow-matching conditioned on history and improves robustness by training on negative-recovery samples with GRPO and a dedicated reasoning dataset.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"vision or high-level guidance while delegating low-level planning to a separate end-to-end module. However, such designs break the end-to-end optimization paradigm, increasing system complexity and training difficulty.2) Only learn- ing from positive samples with limited robustness.Current VLA models only rely on imitation learning from positive/expert trajectories [45,54,75], lead- ing to limited robustness, especially for unseen and long-tail scenarios. However, SpanVLA 3 real-world negative and takeover data, which capture negative behaviors that must be avoided, as well as recovery behavior from challenge scenarios, are often overlooked in datasets and models [66]. Such negative-recovery data can provide targeted refinement signals, improving both performance and robustness."},{"citing_arxiv_id":"2604.17841","ref_index":13,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Driving risk emerges from the required two-dimensional joint evasive acceleration","primary_cat":"cs.RO","submitted_at":"2026-04-20T05:51:18+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Evasive acceleration quantifies driving risk as the minimum 2D constant relative acceleration needed to avoid collision and outperforms time-to-collision on warning timing, discrimination, and information retention across crash datasets.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.15308","ref_index":54,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"RAD-2: Scaling Reinforcement Learning in a Generator-Discriminator Framework","primary_cat":"cs.CV","submitted_at":"2026-04-16T17:59:44+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"RAD-2 uses a diffusion generator and RL discriminator to cut collision rates by 56% in closed-loop autonomous driving planning.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"2 Related Work 2.1 Discriminator for Autonomous Driving Trajectory scoring and selection have emerged as pivotal techniques for enhancing the reliability of autonomous driving systems [9, 36, 46, 63, 67]. Early works like V ADv2 [2] and Hydra-MDP [24, 30] rely on predefined trajectory vocabularies or rule-based teachers to guide se- lection. DriveSuprim [54] further refines this paradigm with a coarse-to-fine filtering framework combined with self-distillation. Recent advances such as DriveDPO [43] and GTRS [33] incorporate preference optimization and dy- namic candidate evaluation to improve flexibility. How- ever, these discriminative approaches typically operate in an open-loop manner, often resulting in suboptimal decisions"},{"citing_arxiv_id":"2604.02714","ref_index":47,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"ExploreVLA: Dense World Modeling and Exploration for End-to-End Autonomous Driving","primary_cat":"cs.CV","submitted_at":"2026-04-03T04:14:13+00:00","verdict":null,"verdict_confidence":null,"novelty_score":null,"formal_verification":null,"one_line_summary":null,"context_count":1,"top_context_role":"baseline","top_context_polarity":"baseline","context_text":"SC: single-view camera; MC: multi-view camera; L: LiDAR;†: best-of-N (N=6) strategy [58]. Model Input NC↑DAC↑EP↑TTC↑Comf.↑ PDMS↑ Ego Status MLP - 93.1 78.3 63.2 84.0 99.9 66.4 TransFuser [5] - 97.8 92.6 78.9 92.9 99.9 83.9 DRAMA [50] MC+L 98.2 95.2 81.3 94.2100.0 86.9 Hydra-MDP [31] MC+L 99.1 98.3 85.2 96.6100.0 91.3 Centaur [36] MC+L 99.2 98.7 86.0 97.2 99.9 92.1 DriveSuprim [47] MC+L 98.6 98.691.395.5100.0 93.5 DrivingGPT [4] SC 98.9 90.7 79.9 94.9 95.6 82.4 FSDrive [52] SC 98.2 93.8 80.1 93.3 99.9 85.1 PWM [54] SC 98.9 95.8 81.5 95.9100.0 88.1 AutoVLA [58] MC 98.4 95.6 81.9 98.0 99.9 89.1 AutoVLA†[58] MC 99.1 97.1 87.6 97.1 99.9 92.1 DriveVLA-W0 [27] SC 98.799.187.6 97.1100.0 90.2 DriveVLA-W0†[27] SC 99.997.4 88.3 97.0 99."},{"citing_arxiv_id":"2604.00813","ref_index":82,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"DVGT-2: Vision-Geometry-Action Model for Autonomous Driving at Scale","primary_cat":"cs.CV","submitted_at":"2026-04-01T12:21:26+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"DVGT-2 is a streaming vision-geometry-action model that jointly reconstructs dense 3D geometry and plans trajectories online, achieving better reconstruction than prior batch methods while transferring directly to planning benchmarks without fine-tuning.","context_count":1,"top_context_role":"baseline","top_context_polarity":"baseline","context_text":"Transfuser [4] C & L Map & Box 97.7 92.8 92.8 100 79.2 84.0 Hydra-MDP [38] C & L Map & Box 98.3 96.0 94.6 100 78.7 86.5 GoalFlow [76] C & L Map & Box 98.3 93.8 94.3 100 79.8 85.7 ARTEMIS [8] C & L Map & Box 98.3 95.1 94.3 100 81.4 87.0 DiffusionDrive [41]C & L Map & Box 98.2 96.2 94.7 100 82.2 88.1 WoTE [35] C & L Map & Box 98.5 96.8 94.9 99.9 81.9 88.3 DriveSuprim [82] C & L Map & Box 97.8 97.3 93.6 100 86.7 89.9 AutoVLA [90] C Language 96.9 92.4 88.1 99.9 75.8 80.5 AdaThinkDrive [53]C Language 98.5 94.4 94.9 100 79.9 86.2 ReCogDrive [37] C Language 98.3 95.1 94.3 100 81.1 86.8 DriveVLA-W0 [34] C Future States 98.7 99.1 95.3 99.3 83.3 90.2 AutoVLA† [90] C Language & RL 98.4 95.6 98.0 99.9 85.9 89.1 ReCogDrive† [37] C Language & RL 98."},{"citing_arxiv_id":"2510.12796","ref_index":31,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"DriveVLA-W0: World Models Amplify Data Scaling Law in Autonomous Driving","primary_cat":"cs.CV","submitted_at":"2025-10-14T17:59:47+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"DriveVLA-W0 adds world modeling to predict future images in VLA models, overcoming sparse action supervision and amplifying data scaling laws on NAVSIM benchmarks and a large in-house dataset.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2507.17596","ref_index":57,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"PRIX: Learning to Plan from Raw Pixels for End-to-End Autonomous Driving","primary_cat":"cs.CV","submitted_at":"2025-07-23T15:28:23+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"PRIX presents an efficient camera-only planner with a novel CaRT module that matches larger multimodal models on NavSim and nuScenes while reducing model size and inference time.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null}],"limit":50,"offset":0}