YILDIZ-VPR: A Novel Dataset with Dense Coverage Under Diverse Environmental Conditions for Visual Place Recognition
为解决视觉地点识别中密集且多样化数据需求的问题，通过在不同环境条件下收集行人视角图像建立了YILDIZ-VPR数据集。
0 citationsRead paper
Selected work
Representative Papers
Deep Agentic Search for Repository-Level Code Question Answering: An Empirical Study
This study investigates whether deep agent search outperforms semantic search in enhancing the accuracy of repository-scale code question answering while mitigating context contamination. We conduct a systematic comparison on the SWE-QA benchmark between vector-index-based semantic search and deep agent search employing sub-agent isolation for exploration. Our analysis reveals, for the first time, that deep agent search suffers from silent errors in the handoff between primary and sub-agents at a rate as high as 41.8%, challenging its prevailing design assumptions. Experimental results demonstrate that semantic search achieves an accuracy of 65.2%, significantly surpassing the 46.2% accuracy of deep agent search, while incurring less than half the cost per response. These findings establish semantic search as superior in both effectiveness and efficiency.
0 citationsRead paper
QEC and EAQEC Codes from Hermitian Sums and Hulls of Cyclic Codes over $\mathbb{F}_2 \times (\mathbb{F}_2+v\mathbb{F}_2)$
This study investigates the construction of high-performance quantum error-correcting (QEC) codes and entanglement-assisted quantum error-correcting (EAQEC) codes from cyclic codes over the composite ring $\mathbb{F}_2 \times (\mathbb{F}_2 + v\mathbb{F}_2)$. By analyzing for the first time the Hermitian hull and Hermitian sum structures of cyclic codes over this ring, the work integrates quantum Construction X, matrix-product codes, and linear complementary dual (LCD) code techniques to propose novel constructions of QEC and EAQEC codes. This approach not only extends the design framework for EAQEC codes but also yields several new families of quantum codes, thereby enriching the theoretical foundation of quantum error correction based on non-traditional algebraic structures.
0 citationsRead paper
Autopilot-Preserving Residual Q-Learning with HJB-Inspired Finite-Action Risk Filtering for Fixed-Wing UAV Command Supervision
This work addresses the challenge of simultaneously tracking airspeed, altitude, and heading for fixed-wing unmanned aerial vehicles under severe wind disturbances, where conventional autopilots exhibit limited adaptability and end-to-end reinforcement learning risks unsafe control surface exploration. The authors propose a learning-based supervisor that preserves the original autopilot architecture by selecting residual corrections to reference commands from a discrete action set, which are then projected onto an admissible command envelope before being fed into the baseline controller—ensuring it remains the sole actuation interface. The approach integrates Hamilton–Jacobi–Bellman (HJB)-inspired principles to formulate an operator-relative advantage scoring mechanism and incorporates a safety filter inspired by control Lyapunov and barrier functions to guarantee safe exploration. Evaluated in a 12-dimensional state space, the method reduces path-tracking RMSE by 86.77% (to 44.8 m) and 49.54% compared to the baseline autopilot and tabular Q-residual methods, respectively, achieving the most significant improvement in scenarios where the original system degrades most severely.
0 citationsRead paper
PYTHALAB-MERA: Validation-Grounded Memory, Retrieval, and Acceptance Control for Frozen-LLM Coding Agents
This work addresses the challenge that local large language models struggle to ensure correctness in coding tasks requiring feedback, persistent state, and limited repair opportunities, due to the absence of verification-driven memory and skill reuse mechanisms. The authors propose a lightweight external controller that operates with frozen model weights and integrates abstract syntax tree (AST)-based skill extraction, a fail-fast verification pipeline, verification-guided memory logging, and TD(λ) eligibility traces for delayed credit assignment. Crucially, verification outcomes are transformed into bounded shaping rewards, enabling structured skill reuse and adaptive retrieval. Evaluated on rigorously verified reinforcement learning coding tasks, the approach succeeds in 8 out of 9 experiments, substantially outperforming both self-refinement baselines and GRACE extensions, which achieve zero successes across all trials.
0 citationsRead paper
Recent publications
Latest Papers
YILDIZ-VPR: A Novel Dataset with Dense Coverage Under Diverse Environmental Conditions for Visual Place Recognition
为解决视觉地点识别中密集且多样化数据需求的问题，通过在不同环境条件下收集行人视角图像建立了YILDIZ-VPR数据集。
0 citationsRead paper
Deep Agentic Search for Repository-Level Code Question Answering: An Empirical Study
This study investigates whether deep agent search outperforms semantic search in enhancing the accuracy of repository-scale code question answering while mitigating context contamination. We conduct a systematic comparison on the SWE-QA benchmark between vector-index-based semantic search and deep agent search employing sub-agent isolation for exploration. Our analysis reveals, for the first time, that deep agent search suffers from silent errors in the handoff between primary and sub-agents at a rate as high as 41.8%, challenging its prevailing design assumptions. Experimental results demonstrate that semantic search achieves an accuracy of 65.2%, significantly surpassing the 46.2% accuracy of deep agent search, while incurring less than half the cost per response. These findings establish semantic search as superior in both effectiveness and efficiency.
0 citationsRead paper
QEC and EAQEC Codes from Hermitian Sums and Hulls of Cyclic Codes over $\mathbb{F}_2 \times (\mathbb{F}_2+v\mathbb{F}_2)$
This study investigates the construction of high-performance quantum error-correcting (QEC) codes and entanglement-assisted quantum error-correcting (EAQEC) codes from cyclic codes over the composite ring $\mathbb{F}_2 \times (\mathbb{F}_2 + v\mathbb{F}_2)$. By analyzing for the first time the Hermitian hull and Hermitian sum structures of cyclic codes over this ring, the work integrates quantum Construction X, matrix-product codes, and linear complementary dual (LCD) code techniques to propose novel constructions of QEC and EAQEC codes. This approach not only extends the design framework for EAQEC codes but also yields several new families of quantum codes, thereby enriching the theoretical foundation of quantum error correction based on non-traditional algebraic structures.
0 citationsRead paper
Autopilot-Preserving Residual Q-Learning with HJB-Inspired Finite-Action Risk Filtering for Fixed-Wing UAV Command Supervision
This work addresses the challenge of simultaneously tracking airspeed, altitude, and heading for fixed-wing unmanned aerial vehicles under severe wind disturbances, where conventional autopilots exhibit limited adaptability and end-to-end reinforcement learning risks unsafe control surface exploration. The authors propose a learning-based supervisor that preserves the original autopilot architecture by selecting residual corrections to reference commands from a discrete action set, which are then projected onto an admissible command envelope before being fed into the baseline controller—ensuring it remains the sole actuation interface. The approach integrates Hamilton–Jacobi–Bellman (HJB)-inspired principles to formulate an operator-relative advantage scoring mechanism and incorporates a safety filter inspired by control Lyapunov and barrier functions to guarantee safe exploration. Evaluated in a 12-dimensional state space, the method reduces path-tracking RMSE by 86.77% (to 44.8 m) and 49.54% compared to the baseline autopilot and tabular Q-residual methods, respectively, achieving the most significant improvement in scenarios where the original system degrades most severely.
0 citationsRead paper
PYTHALAB-MERA: Validation-Grounded Memory, Retrieval, and Acceptance Control for Frozen-LLM Coding Agents
This work addresses the challenge that local large language models struggle to ensure correctness in coding tasks requiring feedback, persistent state, and limited repair opportunities, due to the absence of verification-driven memory and skill reuse mechanisms. The authors propose a lightweight external controller that operates with frozen model weights and integrates abstract syntax tree (AST)-based skill extraction, a fail-fast verification pipeline, verification-guided memory logging, and TD(λ) eligibility traces for delayed credit assignment. Crucially, verification outcomes are transformed into bounded shaping rewards, enabling structured skill reuse and adaptive retrieval. Evaluated on rigorously verified reinforcement learning coding tasks, the approach succeeds in 8 out of 9 experiments, substantially outperforming both self-refinement baselines and GRACE extensions, which achieve zero successes across all trials.
0 citationsRead paper