AgentLens: Revealing The Lucky Pass Problem in SWE-Agent Evaluation

Evaluation of software engineering (SWE) agents is dominated by a binary signal: whether the final patch passes the tests. This outcome-only view treats a principled solution and a chaotic trial-and-error process as equivalent. We show that this equivalence is empirically false. We evaluate 2,614 OpenHands trajectories from eight model backends on 60 SWE-bench Verified tasks. Of these, 47 have enough passing trajectories to construct task-level process references, yielding a 1,815-trajectory evaluation subset. Among passing trajectories in this subset, 10.7% exhibit behavior we call a Lucky Pass: regression cycles, blind retries, missing verification, or temporally disordered exploration, implementation, and verification. We introduce AgentLens, a framework for process-level assessment of SWE-agent trajectories, and release AgentLens-Bench, a dataset of 1,815 trajectories annotated with quality scores, waste signals, divergence points, and 47 task-level Prefix Tree Acceptor (PTA) references. AgentLens builds PTA references by merging multiple passing solutions for the same task, and uses a context-sensitive intent labeler to assign actions to Exploration, Implementation, Verification, or Orchestration based on trajectory history rather than tool identity alone. On AgentLens-Bench, the quality score separates passing trajectories into Lucky, Solid, and Ideal tiers and further decomposes Lucky Passes into five recurring mechanisms. Across the eight model backends, Lucky rates range from 0.5% to 23.2%, and some models move by as many as five rank positions when ranked by quality score instead of pass rate. We release the anonymized project repository, including the AgentLens-Bench dataset and AgentLens SDK, at https://github.com/microsoft/code-agent-state-trajectories/.

Multi-Rollout On-Policy Distillation via Peer Successes and Failures

Large language models are often post-trained with sparse verifier rewards, which indicate whether a sampled trajectory succeeds but provide limited guidance about where reasoning succeeds or fails. On-policy distillation (OPD) offers denser token-level supervision by training on student-generated trajectories, yet existing methods typically distill each rollout independently and ignore the other attempts sampled for the same prompt. We introduce Multi-Rollout On-Policy Distillation (MOPD), a peer-conditioned distillation framework that uses the student's local rollout group to construct more informative teacher signals. MOPD conditions the teacher on both successful and failed peer rollouts: successes provide positive evidence for valid reasoning patterns, while failures provide structured negative evidence about plausible mistakes to avoid. We study two peer-context constructions: positive peer imitation and contrastive success-failure conditioning. Experiments on competitive programming, mathematical reasoning, scientific question answering, and tool-use benchmarks show that MOPD consistently improves over standard on-policy baselines. Further teacher-signal analysis shows that mixed success-failure contexts better align teacher scores with verifier rewards, indicating that the gains arise from more faithful, instance-adaptive supervision. These results indicate that effective on-policy distillation should exploit the student's multi-rollout trial-and-error behavior rather than treating rollouts as isolated samples.

Ground4D: Spatially-Grounded Feedforward 4D Reconstruction for Unstructured Off-Road Scenes

Feedforward Gaussian Splatting has recently emerged as an efficient paradigm for 4D reconstruction in autonomous driving. However, in unstructured off-road scenes, its performance degrades due to high-frequency geometry, ego-motion jitter, and increased non-rigid dynamics. These factors introduce conflicting Gaussian observations across timestamps, leading to either over-smoothed renderings or structural artifacts. To address this issue, we propose Ground4D, a spatially-grounded 4D feedforward framework for pose-free off-road reconstruction. The key idea is to resolve temporal conflicts through spatially localized conditioning. Specifically, we introduce voxel-grounded temporal Gaussian aggregation, which partitions the canonical Gaussian space into spatial voxels and performs query-conditioned temporal attention within each voxel. Intra-voxel softmax normalization ensures that temporal selectivity and spatial occupancy become mutually reinforcing rather than conflicting. We furthermore introduce surface normal cues as auxiliary geometric guidance to regularize the geometry of Gaussian primitives. Extensive experiments on ORAD-3D and RELLIS-3D demonstrate that Ground4D consistently outperforms existing feedforward methods in reconstruction quality and generalizes zero-shot to unseen off-road domains. Project page and code:https://github.com/wsnbws/Ground4D.

Learning Correct Behavior from Examples: Validating Sequential Execution in Autonomous Agents

As autonomous agents become increasingly sophisticated, validating their sequential behavior presents a significant challenge. Traditional testing approaches require manual specification, exact sequence matching, or thousands of training examples. We present a novel algorithm that automatically learns correct behavior from just 2-10 passing execution traces and validates new executions against this learned model. Our approach combines dominator analysis from compiler theory with multimodal large language model-powered semantic understanding to identify essential states and handle non-deterministic behavior. The system constructs a generalized ground truth model using Prefix Tree Acceptors, merges traces through multi-tiered equivalence detection, and validates new executions via topological subsequence matching. In controlled experiments, our system achieved high accuracy in detecting product bugs and false successes using only 3 training traces. This approach provides explainable validation results with coverage metrics and works across diverse domains including UI testing, code generation, and robotic processes.

See, Point, Refine: Multi-Turn Approach to GUI Grounding with Visual Feedback

Computer Use Agents (CUAs) fundamentally rely on graphical user interface (GUI) grounding to translate language instructions into executable screen actions, but editing-level grounding in dense coding interfaces, where sub-pixel accuracy is required to interact with dense IDE elements, remains underexplored. Existing approaches typically rely on single-shot coordinate prediction, which lacks a mechanism for error correction and often fails in high-density interfaces. In this technical report, we conduct an empirical study of pixel-precise cursor localization in coding environments. Instead of a single-step execution, our agent engages in an iterative refinement process, utilizing visual feedback from previous attempts to reach the target element. This closed-loop grounding mechanism allows the agent to self-correct displacement errors and adapt to dynamic UI changes. We evaluate our approach across GPT-5.4, Claude, and Qwen on a suite of complex coding benchmarks, demonstrating that multi-turn refinement significantly outperforms state-of-the-art single-shot models in both click precision and overall task success rate. Our results suggest that iterative visual reasoning is a critical component for the next generation of reliable software engineering agents. Code: https://github.com/microsoft/precision-cua-bench.

E2E-Fly: An Integrated Training-to-Deployment System for End-to-End Quadrotor Autonomy

Training and transferring learning-based policies for quadrotors from simulation to reality remains challenging due to inefficient visual rendering, physical modeling inaccuracies, unmodeled sensor discrepancies, and the absence of a unified platform integrating differentiable physics learning into end-to-end training. While recent work has demonstrated various end-to-end quadrotor control tasks, few systems provide a systematic, zero-shot transfer pipeline, hindering reproducibility and real-world deployment. To bridge this gap, we introduce E2E-Fly, an integrated framework featuring an agile quadrotor platform coupled with a full-stack training, validation, and deployment workflow. The training framework incorporates a high-performance simulator with support for differentiable physics learning and reinforcement learning, alongside structured reward design tailored to common quadrotor tasks. We further introduce a two-stage validation strategy using sim-to-sim transfer and hardware-in-the-loop testing, and deploy policies onto two physical quadrotor platforms via a dedicated low-level control interface and a comprehensive sim-to-real alignment methodology, encompassing system identification, domain randomization, latency compensation, and noise modeling. To the best of our knowledge, this is the first work to systematically unify differentiable physical learning with training, validation, and real-world deployment for quadrotors. Finally, we demonstrate the effectiveness of our framework for training six end-to-end control tasks and deploy them in the real world.

Vision-Based End-to-End Learning for UAV Traversal of Irregular Gaps via Differentiable Simulation

-Navigation through narrow and irregular gaps is an essential skill in autonomous drones for applications such as inspection, search-and-rescue, and disaster response. However, traditional planning and control methods rely on explicit gap extraction and measurement, while recent end-to-end approaches often assume regularly shaped gaps, leading to poor generalization and limited practicality. In this work, we present a fully vision-based, end-to-end framework that maps depth images directly to control commands, enabling drones to traverse complex gaps within unseen environments. Operating in the Special Euclidean group SE(3), where position and orientation are tightly coupled, the framework leverages differentiable simulation, a Stop-Gradient operator, and a Bimodal Initialization Distribution to achieve stable traversal through consecutive gaps. Two auxiliary prediction modules-a gap-crossing success classifier and a traversability predictor-further enhance continuous navigation and safety. Extensive simulation and real-world experiments demonstrate the approach's effectiveness, generalization capability, and practical robustness.

AVION: Aerial Vision-Language Instruction from Offline Teacher to Prompt-Tuned Network

Adapting vision-language models to remote sensing imagery remains challenging due to two key factors: limited semantic coverage in textual representations and insufficient adaptability of visual features. These issues are particularly significant in aerial scenes, which involve various visual appearances and fine-grained object distinctions. We propose AVION, a knowledge distillation framework tailored for remote sensing adaptation of vision-language models. The teacher module constructs semantically rich textual prototypes by collecting descriptions from a large language model and verifying validity using remote sensing image features. The student module integrates lightweight and learnable prompts into both vision and language encoders, guided by the teacher to align embeddings and their cross-modal relationships. Once trained, the student operates independently during inference. Experiments on six optical remote sensing benchmarks show that AVION improves few-shot classification and base-class accuracy without degrading generalization to novel categories. It also enhances mean recall for cross-modal retrieval, with minimal additional trainable parameters.

Vector Field Augmented Differentiable Policy Learning for Vision-Based Drone Racing

Autonomous drone racing in complex environments requires agile, high-speed flight while maintaining reliable obstacle avoidance. Differentiable-physics-based policy learning has recently demonstrated high sample efficiency and remarkable performance across various tasks, including agile drone flight and quadruped locomotion. However, applying such methods to drone racing remains difficult, as key objective like gate traversal are inherently hard to express as smooth, differentiable losses. To address these challenges, we propose DiffRacing, a novel vector field-augmented differentiable policy learning framework. DiffRacing integrates differentiable losses and vector fields into the training process to provide continuous and stable gradient signals, balancing obstacle avoidance and high-speed gate traversal. In addition, a differentiable Delta Action Model compensates for dynamics mismatch, enabling efficient sim-to-real transfer without explicit system identification. Extensive simulation and real-world experiments demonstrate that DiffRacing achieves superior sample efficiency, faster convergence, and robust flight performance, thereby demonstrating that vector fields can augment traditional gradient-based policy learning with a task-specific geometric prior.

AI-IO: An Aerodynamics-Inspired Real-Time Inertial Odometry for Quadrotors

Inertial Odometry (IO) has gained attention in quadrotor applications due to its sole reliance on inertial measurement units (IMUs), attributed to its lightweight design, low cost, and robust performance across diverse environments. However, most existing learning-based inertial odometry systems for quadrotors either use only IMU data or include additional dynamics-related inputs such as thrust, but still lack a principled formulation of the underlying physical model to be learned. This lack of interpretability hampers the model's ability to generalize and often limits its accuracy. In this work, we approach the inertial odometry learning problem from a different perspective. Inspired by the aerodynamics model and IMU measurement model, we identify the key physical quantity--rotor speed measurements required for inertial odometry and design a transformer-based inertial odometry. By incorporating rotor speed measurements, the proposed model improves velocity prediction accuracy by 36.9%. Furthermore, the transformer architecture more effectively exploits temporal dependencies for denoising and aerodynamics modeling, yielding an additional 22.4% accuracy gain over previous results. To support evaluation, we also provide a real-world quadrotor flight dataset capturing IMU measurements and rotor speed for high-speed motion. Finally, combined with an uncertainty-aware extended Kalman filter (EKF), our framework is validated across multiple datasets and real-time systems, demonstrating superior accuracy, generalization, and real-time performance. We share the code and data to promote further research (https://github.com/SJTU-ViSYS-team/AI-IO).