Searcharxiv⌕ Search

arXiv subjects

Rongqing Li

Publications and source records attributed to Rongqing Li.

8 recordsLinked to original sources

Coding Agent Memory Post-Training II: Unlocking the Memory Potential of Pre-trained File Operations for Long-Horizon Tasks via On-Policy Distillation

Language-model agents increasingly tackle long-horizon tasks whose interaction histories exceed the model's active context. They must therefore decide what to write down and what to read back, and when to do either. Recent work has begun to use on-policy distillation (OPD) to make memory control part of the policy, relying on memory tools and compaction protocols that lie outside either model's pre-training. However, this weakens the distillation on both sides: the teacher cannot give useful guidance on an interface it has never used, and the student samples too few meaningful memory actions for that guidance to take effect. To study this problem, we introduce Coding Agent Memory Gym II (CAMG-II), a suite of long-horizon agentic-RL environments spanning Shop, Coding, AutoResearch, General, and Visual. Alongside each environment's native task interface, CAMG-II provides ordinary Bash access and an episode-persistent workspace, enabling the policy to create, revise, search, and reuse files as memory across context replacements. We also introduce Coding Agent Memory Post-Training II (CAMP-II), which trains a single policy jointly across all five environments with fully asynchronous OPD to learn this file-based memory behavior directly from a frozen teacher with a prior for file operations. On CAMG-II, CAMP-II obtains 2.97 times the gain of the strongest OPD baseline. The memory behavior CAMP-II learns generalizes to two external benchmarks, Terminal-Bench 2.1 and SWE-Bench Pro V2, with improved success rates.

cs.LG↗

SWE-Shift: Evaluating Coding Agents in Evolving Workspaces

Repository-level coding benchmarks largely assume that tasks and environments remain unchanged throughout execution. In software development, code, dependencies, runtime state, and requirements can change while an agent is still working. We introduce SWE-Shift, a benchmark for evaluating coding agents under such mid-execution changes. From the same saved history and workspace, a Clean continuation proceeds without intervention, while a matched Shifted continuation receives a task-relevant update. Success requires completing the current task, preserving still-valid work, and satisfying the applicable state and evidence checks. The benchmark contains 120 scenarios across 12 change families, each evaluated at nine scheduled intervention positions. We evaluate 17 coding-agent systems, all of which have lower Shifted than Clean success. GPT-6 Astra drops from 67.35% to 24.79%, with 20.34% paired success. Tool/session changes and stale evidence yield zero paired success across all systems. Late-position paired success is lower for some systems and higher for others. These results show that success on the unchanged task can coexist with failures to preserve work, recover state, or refresh evidence after an update.

cs.AI↗

Coding Agent Memory Post-training: Unlocking the Memory Potential of Pre-trained File Operations for Long-Horizon Tasks via Reinforcement Learning

Language-model agents increasingly tackle long-horizon tasks whose interaction histories exceed the model's active context. Recent work has begun to use reinforcement learning to make memory control part of the policy, often relying on predefined memory tools within domain-specific training environments of relatively short horizons. This setup ties learned memory behavior to environment-specific interfaces that lie outside the base model's pre-training and must be learned from scratch, so even after post-training, agents struggle to use memory in long-horizon tasks. To address these limitations, we introduce Coding Agent Memory Gym (CAMG), a suite of long-horizon agentic-RL environments spanning Shop, Coding, DeepResearch, and AutoResearch. Alongside each environment's native task interface, CAMG provides executable shell access and an episode-persistent workspace, enabling agents to create, revise, search, and reuse files as memory throughout an episode. We also introduce CAMG-RL, which trains a single policy jointly across all four environments with fully asynchronous PPO, learning this file-based memory behavior directly from downstream task reward, and we train CAMG-RL-4B and CAMG-RL-9B from Qwen3.5 models of matching size. On SWE-bench Verified and MLE-bench Lite, CAMG-RL-4B and CAMG-RL-9B are competitive with Qwen3.5-35B-A3B and Qwen3.5-122B-A10B, respectively.

cs.LG↗

Beyond Forgetting: Diagnosing and Harnessing Shared Reasoning in Continual RLVR

Reinforcement learning with verifiable rewards (RLVR) commonly post-trains reasoning models on multiple tasks, while rerunning multitask RLVR (MTRL) as new tasks are added makes capability expansion costly. We therefore study continual RLVR, which updates the existing model as each task arrives. The central question is whether a model updated this way can perform as well as a jointly trained model. To answer this question, we introduce Continual Reasoning Gym, a continual-RLVR environment that organizes text and visual reasoning tasks into five task sequences. In this setting, we identify two key observations: Sequential RLVR exhibits modest forgetting, yet its final performance remains below that of MTRL. To understand the latter, we decompose final performance and show that forgetting accounts for only part of the gap. To explain the former, we identify shared reasoning: transferable reasoning structure allows training on one task to support others on average. We therefore introduce Continual Prompt Replay (CPR), which harnesses shared reasoning to improve learning on the arriving and future tasks by replaying previous-task prompts and regenerating their responses with the current policy. On average, only CPR reaches MTRL-level performance.

cs.LG↗

ITPNet: Towards Instantaneous Trajectory Prediction for Autonomous Driving

Trajectory prediction of agents is crucial for the safety of autonomous vehicles, whereas previous approaches usually rely on sufficiently long-observed trajectory to predict the future trajectory of the agents. However, in real-world scenarios, it is not realistic to collect adequate observed locations for moving agents, leading to the collapse of most prediction models. For instance, when a moving car suddenly appears and is very close to an autonomous vehicle because of the obstruction, it is quite necessary for the autonomous vehicle to quickly and accurately predict the future trajectories of the car with limited observed trajectory locations. In light of this, we focus on investigating the task of instantaneous trajectory prediction, i.e., two observed locations are available during inference. To this end, we propose a general and plug-and-play instantaneous trajectory prediction approach, called ITPNet. Specifically, we propose a backward forecasting mechanism to reversely predict the latent feature representations of unobserved historical trajectories of the agent based on its two observed locations and then leverage them as complementary information for future trajectory prediction. Meanwhile, due to the inevitable existence of noise and redundancy in the predicted latent feature representations, we further devise a Noise Redundancy Reduction Former, aiming at to filter out noise and redundancy from unobserved trajectories and integrate the filtered features and observed features into a compact query for future trajectory predictions. In essence, ITPNet can be naturally compatible with existing trajectory prediction models, enabling them to gracefully handle the case of instantaneous trajectory prediction. Extensive experiments on the Argoverse and nuScenes datasets demonstrate ITPNet outperforms the baselines, and its efficacy with different trajectory prediction models.

cs.CV↗

DREAM: Domain-agnostic Reverse Engineering Attributes of Black-box Model

Deep learning models are usually black boxes when deployed on machine learning platforms. Prior works have shown that the attributes (e.g., the number of convolutional layers) of a target black-box model can be exposed through a sequence of queries. There is a crucial limitation: these works assume the training dataset of the target model is known beforehand and leverage this dataset for model attribute attack. However, it is difficult to access the training dataset of the target black-box model in reality. Therefore, whether the attributes of a target black-box model could be still revealed in this case is doubtful. In this paper, we investigate a new problem of black-box reverse engineering, without requiring the availability of the target model's training dataset. We put forward a general and principled framework DREAM, by casting this problem as out-of-distribution (OOD) generalization. In this way, we can learn a domain-agnostic meta-model to infer the attributes of the target black-box model with unknown training data. This makes our method one of the kinds that can gracefully apply to an arbitrary domain for model attribute reverse engineering with strong generalization ability. Extensive experimental results demonstrate the superiority of our proposed method over the baselines.

cs.LG↗

DREAM: Domain-free Reverse Engineering Attributes of Black-box Model

Deep learning models are usually black boxes when deployed on machine learning platforms. Prior works have shown that the attributes ($e.g.$, the number of convolutional layers) of a target black-box neural network can be exposed through a sequence of queries. There is a crucial limitation: these works assume the dataset used for training the target model to be known beforehand and leverage this dataset for model attribute attack. However, it is difficult to access the training dataset of the target black-box model in reality. Therefore, whether the attributes of a target black-box model could be still revealed in this case is doubtful. In this paper, we investigate a new problem of Domain-agnostic Reverse Engineering the Attributes of a black-box target Model, called DREAM, without requiring the availability of the target model's training dataset, and put forward a general and principled framework by casting this problem as an out of distribution (OOD) generalization problem. In this way, we can learn a domain-agnostic model to inversely infer the attributes of a target black-box model with unknown training data. This makes our method one of the kinds that can gracefully apply to an arbitrary domain for model attribute reverse engineering with strong generalization ability. Extensive experimental studies are conducted and the results validate the superiority of our proposed method over the baselines.

cs.LG↗

Deep Dense Multi-scale Network for Snow Removal Using Semantic and Geometric Priors

Images captured in snowy days suffer from noticeable degradation of scene visibility, which degenerates the performance of current vision-based intelligent systems. Removing snow from images thus is an important topic in computer vision. In this paper, we propose a Deep Dense Multi-Scale Network (\textbf{DDMSNet}) for snow removal by exploiting semantic and geometric priors. As images captured in outdoor often share similar scenes and their visibility varies with depth from camera, such semantic and geometric information provides a strong prior for snowy image restoration. We incorporate the semantic and geometric maps as input and learn the semantic-aware and geometry-aware representation to remove snow. In particular, we first create a coarse network to remove snow from the input images. Then, the coarsely desnowed images are fed into another network to obtain the semantic and geometric labels. Finally, we design a DDMSNet to learn semantic-aware and geometry-aware representation via a self-attention mechanism to produce the final clean images. Experiments evaluated on public synthetic and real-world snowy images verify the superiority of the proposed method, offering better results both quantitatively and qualitatively.

cs.CV↗