A Verifier-Guided Explainable Reasoning Framework with Gold-Anchored QLoRA, Task-Aware Mixture-of-Experts, and Group-Relative RLVR
本文提出一种验证器引导的可解释推理框架，通过黄金锚定QLoRA、任务感知混合专家系统和组相对RLVR方法，提高大语言模型在教育问答中的解释性和准确性。
0 citationsRead paper
Selected work
Representative Papers
A Revisit to Point Estimation Through the Empirical Bayes Method: The Case of Binomial Distribution with Beta Prior and Extension to Poisson Distribution
This study reevaluates the efficacy of empirical Bayes methods for parameter estimation in binomial (with Beta priors) and Poisson models. Through theoretical analysis and extensive numerical experiments, it specifically investigates Type-II maximum likelihood (ML-II) under general two-parameter Beta priors and extends the examination to the Gamma–Poisson setting. The findings reveal that the ML-II procedure fails under a general two-parameter Beta prior; even when restricted to a symmetric one-parameter Beta prior, the resulting estimator does not substantially outperform the maximum likelihood estimator under quadratic loss. These results challenge the commonly presumed superiority of empirical Bayes approaches in point estimation and provide a critical counterexample grounded in rigorous empirical evidence.
0 citationsRead paper
FaLCon: Facet-Anchored Retrieval with Late Consensus for Sim2Real Text-Based Person Anomaly Search
This work addresses the challenge of fine-grained matching in text-based pedestrian anomaly retrieval under synthetic-to-real (Sim2Real) scenarios. To this end, the authors propose an anchor-constrained coarse-to-fine retrieval framework that leverages multi-facet semantic decomposition and calibrated fusion. The approach integrates a heterogeneous vision-language retriever, a Qwen3-based reranker, and an anomaly-aware cloze-style verification module, complemented by an uncertainty-gated consensus mechanism operating over a small candidate pool to enable efficient fine-grained semantic reasoning. Innovatively, semantic facets serve as anchor constraints to jointly optimize recall and computational efficiency. Evaluated on the PAB benchmark, the method achieves 95.41% mAP@10, 94.44% R@1, and 99.09% R@5, significantly outperforming existing single-backbone models.
0 citationsRead paper
GeoRoute: Geometry-Aware Hybrid Inference for Traffic Future-Frame Prediction
Long-term traffic scene prediction often suffers from temporal ghosting, geometric drift, and motion inconsistency, leading to unstable static structures and temporally incoherent outputs. To address this, this work proposes a training-free inference framework that enhances the outputs of pretrained video diffusion models through geometry-aware refinement during inference. The approach uniquely integrates multi-frame depth-layered rendering with vision-language model–guided view-conditioned routing, leveraging a frozen visual language model to achieve geometric stability and cross-view generalization without fine-tuning. Evaluated on the AI City Challenge Track 5, the method achieves state-of-the-art performance, significantly improving fidelity of static structures and consistency of low-level details.
0 citationsRead paper
Causal Episodic Memory for Feedback-Driven Agent Repair
This work addresses the tendency of large language model (LLM) agents to discard previously validated successful corrections during task repair, leading to redundant exploration. To mitigate this, the authors propose MERIT, a training-free agent that introduces, for the first time, a causality-aware, type-guided bipolar memory mechanism to explicitly distinguish between successful and failed repair trajectories. Built upon a frozen LLM, MERIT enables lightweight cross-query experience reuse through a deterministic failure classifier and a hybrid lexical-dense retriever. Evaluated on the Spider and BIRD datasets, MERIT improves execution accuracy from 66.34% to 69.79% and from 47.35% to 48.44%, respectively, demonstrating the efficacy of its causal memory architecture.
0 citationsRead paper
Recent publications
Latest Papers
A Verifier-Guided Explainable Reasoning Framework with Gold-Anchored QLoRA, Task-Aware Mixture-of-Experts, and Group-Relative RLVR
本文提出一种验证器引导的可解释推理框架，通过黄金锚定QLoRA、任务感知混合专家系统和组相对RLVR方法，提高大语言模型在教育问答中的解释性和准确性。
0 citationsRead paper
A Revisit to Point Estimation Through the Empirical Bayes Method: The Case of Binomial Distribution with Beta Prior and Extension to Poisson Distribution
This study reevaluates the efficacy of empirical Bayes methods for parameter estimation in binomial (with Beta priors) and Poisson models. Through theoretical analysis and extensive numerical experiments, it specifically investigates Type-II maximum likelihood (ML-II) under general two-parameter Beta priors and extends the examination to the Gamma–Poisson setting. The findings reveal that the ML-II procedure fails under a general two-parameter Beta prior; even when restricted to a symmetric one-parameter Beta prior, the resulting estimator does not substantially outperform the maximum likelihood estimator under quadratic loss. These results challenge the commonly presumed superiority of empirical Bayes approaches in point estimation and provide a critical counterexample grounded in rigorous empirical evidence.
0 citationsRead paper
FaLCon: Facet-Anchored Retrieval with Late Consensus for Sim2Real Text-Based Person Anomaly Search
This work addresses the challenge of fine-grained matching in text-based pedestrian anomaly retrieval under synthetic-to-real (Sim2Real) scenarios. To this end, the authors propose an anchor-constrained coarse-to-fine retrieval framework that leverages multi-facet semantic decomposition and calibrated fusion. The approach integrates a heterogeneous vision-language retriever, a Qwen3-based reranker, and an anomaly-aware cloze-style verification module, complemented by an uncertainty-gated consensus mechanism operating over a small candidate pool to enable efficient fine-grained semantic reasoning. Innovatively, semantic facets serve as anchor constraints to jointly optimize recall and computational efficiency. Evaluated on the PAB benchmark, the method achieves 95.41% mAP@10, 94.44% R@1, and 99.09% R@5, significantly outperforming existing single-backbone models.
0 citationsRead paper
GeoRoute: Geometry-Aware Hybrid Inference for Traffic Future-Frame Prediction
Long-term traffic scene prediction often suffers from temporal ghosting, geometric drift, and motion inconsistency, leading to unstable static structures and temporally incoherent outputs. To address this, this work proposes a training-free inference framework that enhances the outputs of pretrained video diffusion models through geometry-aware refinement during inference. The approach uniquely integrates multi-frame depth-layered rendering with vision-language model–guided view-conditioned routing, leveraging a frozen visual language model to achieve geometric stability and cross-view generalization without fine-tuning. Evaluated on the AI City Challenge Track 5, the method achieves state-of-the-art performance, significantly improving fidelity of static structures and consistency of low-level details.
0 citationsRead paper
Causal Episodic Memory for Feedback-Driven Agent Repair
This work addresses the tendency of large language model (LLM) agents to discard previously validated successful corrections during task repair, leading to redundant exploration. To mitigate this, the authors propose MERIT, a training-free agent that introduces, for the first time, a causality-aware, type-guided bipolar memory mechanism to explicitly distinguish between successful and failed repair trajectories. Built upon a frozen LLM, MERIT enables lightweight cross-query experience reuse through a deterministic failure classifier and a hybrid lexical-dense retriever. Evaluated on the Spider and BIRD datasets, MERIT improves execution accuracy from 66.34% to 69.79% and from 47.35% to 48.44%, respectively, demonstrating the efficacy of its causal memory architecture.
0 citationsRead paper