Searcharxiv⌕ Search

arXiv subjects

Wenzheng Kong

Publications and source records attributed to Wenzheng Kong.

3 recordsLinked to original sources

A Full-Ledger Counterfactual Audit of Adaptive Reasoning

Adaptive reasoning must decide whether to answer now or spend more compute. Predicting that the current answer is wrong does not establish whether further reasoning will improve it. We introduce the Full-Ledger Counterfactual Value Audit (FLCVA), which compares an immediate answer with four fixed-length continuations from the same frozen prefix under one answer procedure and accounts for the cost of every signal used by a policy. On GSM8K and MATH-500, a random 256-token continuation recovers a wrong answer in 26.6% of Qwen3-8B cases and 18.9% of DeepSeek-R1-Distill-Qwen-7B cases, while damaging a correct answer in 1.4% and 1.9%. Correctness labels from independently sampled continuations predict a separate continuation's outcome in both primary models. In offline replay, using these labels to allocate compute improves performance on DeepSeek; the Qwen3 gain remains inconclusive. Under full cost accounting, the tested lightweight, myopic controllers mainly recognize current errors and add little continuation-specific value beyond that shortcut. A Llama-8B replication and action/cost ablations examine the robustness of these findings. FLCVA separates measuring local continuation value, predicting it from observable signals, and using those predictions to allocate computation.

cs.AI↗

G-CAST: Governed Continual Self-Training for E-Commerce Agents under Evolving Demands

With the rapid advancement of Large Language Models (LLMs), an increasing number of users are turning to LLM-powered agents for conversational shopping in e-commerce. However, the user needs continue to change as new product categories, procurement scenarios, and feature requests emerge. At the same time, the product's execution harness-including its tools, skills, and operating procedures-continues to evolve to support increasingly sophisticated interactions. These changes create two practical challenges: 1) Demand drift alters the distribution, intent, and complexity of user requests, causing a model trained once to fall behind; 2) Harness evolution requires the model to use newly introduced capabilities correctly without losing behavior that users already rely on. Periodic manual retraining is too delayed and labor-intensive for this setting. To address these challenges, we present G-CAST, a governed continual self-training system that converts daily production evidence under evolving demand and capability contracts into reward-routed, entropy-prioritized training data and governed reward evolution. Governed checkpoint transition under cumulative evaluation controls each model update, enabling the agent to learn newly introduced behaviors without sacrificing previously accepted capabilities. Experiments on a real production e-commerce agent show that G-CAST achieves stronger adaptation to newly introduced demand while better preserving previously accepted behavior compared with a strong autonomous post-training baseline. Ablations further show that both reward-routed, entropy-prioritized data orchestration and governed reward evolution are important to these gains. G-CAST is now deployed in Yuancai(http://cai.1688.com/), a B2B procurement agent on Alibaba's 1688 platform.

cs.AI↗

Learning-Based Interface for Semantic Communication with Bit Importance Awareness

Joint source-channel coding (JSCC) is an effective approach for semantic communication. However, current JSCC methods are difficult to integrate with existing communication network architectures, where application and network providers are typically different entities. Recently, a novel paradigm termed Split DeepJSCC has been under consideration to address this challenge. Split DeepJSCC employs a bit-level interface that enables separate design of source and channel codes, ensuring compatibility with existing communication networks while preserving the advantages of JSCC in terms of semantic fidelity and channel adaptability. In this paper, we propose a learning-based interface design by treating its parameters as trainable, achieving improved end-to-end performance compared to Split DeepJSCC. In particular, the interface enables specification of bit-level importance at the output of the source code. Furthermore, we propose an Importance-Aware Net that utilizes the interface-derived bit importance information, enabling dynamical adaptation to diverse channel bandwidth ratios and time-varying channel conditions. Experimental results show that our method improves performance in wireless image transmission tasks. This work provides a potential solution for realizing semantic communications in existing wireless networks.

cs.IT↗