Zhijie Sun

Zhicheng Zhou, Jing Li, Suming Qiu et al.

General-purpose large language models (LLMs) are increasingly deployed in verticals such as telecommunications, where adaptation is hindered by scarce, low-information-density corpora and tight mobile/edge constraints. We propose Data Trajectory Alignment (DTA), a two-phase, model-agnostic data curation framework that treats solution processes - not only final answers - as first-class supervision. Phase I (Initializing) synthesizes diverse, high-coverage candidates using an ensemble of strong teachers. Phase II (DTA) rewrites teacher solutions to align intermediate steps and presentation style with the target student's inductive biases and then performs signal-aware exemplar selection via agreement checks and reflection-based judging. Instantiated on telecommunications mathematics (e.g., link budgets, SNR/AMC selection, and power-control feasibility), DTA yields state-of-the-art (SOTA) accuracy on TELEMATH without enabling explicit "thinking" modes: 72.45% pass@1, surpassing distilled-only training by +17.65 points and outperforming a strong baseline (Qwen3-32B with thinking enabled) by +2.94 points. Token-shift analyses indicate that DTA concentrates gains on logical-structural discourse markers rather than merely amplifying domain nouns, indicating improved reasoning scaffolding. Under edge-like inference settings, DTA improves efficiency by reducing reliance on multi-sample voting and disabling expensive reasoning heuristics, cutting energy per output token by ~42% versus Qwen3-32B (thinking mode enabled) and end-to-end latency by ~60% versus Qwen3-32B (thinking mode disabled). These results demonstrate that aligning how solutions are produced enables compact, high-yield supervision that is effective for both accuracy and efficiency, offering a practical recipe for domain adaptation in low-resource verticals beyond telecom.

Zhicheng Zhou, Jing Li, Suming Qiu et al.

The internet is saturated with low-density, high-redundancy information, such as social media comments, repetitive news, and lengthy discussions, making it difficult to extract valuable insights efficiently. Multi-layer nested JSON structures provide an effective solution by compressing such information into semantically rich, hierarchical representations, which organize data into key-value pairs, arrays, and nested objects, preserving contextual relationships and enabling efficient storage, retrieval, and semantic querying. For instance, in news aggregation, a JSON object can nest an article's metadata (title, author, date), content (text, multimedia), and multimedia information (multimedia type, caption) hierarchically. Large Language Models (LLMs) play a transformative role in web data mining by parsing unstructured text and outputting structured results directly into complex JSON schemas. However, current benchmarks for evaluating LLMs' JSON output capabilities overemphasize pure JSON generation rather than assessing data comprehension and extraction abilities, a limitation that lacks relevance to practical web data mining tasks. To address this, we introduce DeepJSONEval, a novel benchmark featuring 2100 multi-domain instances with deep nested structures, categorized by difficulty. Experiments show significant performance gaps among LLMs in handling such complexity. Our benchmark and datasets are open-sourced to advance research in structured JSON generation.(https://github.com/GTS-AI-Infra-Lab-SotaS/DeepJSONEval).

Jing Li, Zhijie Sun, Zhicheng Zhou et al.

Current knowledge-enhanced large language models (LLMs) rely on static, pre-constructed knowledge bases that suffer from coverage gaps and temporal obsolescence, limiting their effectiveness in dynamic information environments. We present Agentic-KGR, a novel framework enabling co-evolution between LLMs and knowledge graphs (KGs) through multi-round reinforcement learning (RL). Our approach introduces three key innovations: (1) a dynamic schema expansion mechanism that systematically extends graph ontologies beyond pre-defined boundaries during training; (2) a retrieval-augmented memory system enabling synergistic co-evolution between model parameters and knowledge structures through continuous optimization; (3) a learnable multi-scale prompt compression approach that preserves critical information while reducing computational complexity through adaptive sequence optimization. Experimental results demonstrate substantial improvements over supervised baselines and single-round RL approaches in knowledge extraction tasks. When integrated with GraphRAG, our method achieves superior performance in downstream QA tasks, with significant gains in both accuracy and knowledge coverage compared to existing methods.

Suming Qiu, Jing Li, Zhicheng Zhou et al.

Large language models (LLMs) often face a trade-off in post-training: improvements on specialized domains frequently come at the expense of general capabilities. Existing solutions attempt to mitigate this tension via regularization, selective parameter updates, or data-centric replay, but each imposes significant costs in computation, data access, or adaptability. Recent work has shown that training signals can be compressed to subsets of logits without severe accuracy loss, suggesting a path toward efficient adaptation. However, naive truncation destabilizes optimization and exacerbates forgetting. We introduce Logits Replay + MoClip, a two-stage framework that compresses supervision in the logit space and stabilizes optimization at the update level. In Stage 0, we record dynamic Top-K token subsets that cover a probability threshold, always including the gold label. In Stage 1, we replay these compact subsets to compute exact renormalized losses, avoiding full softmax computation and implicitly regularizing. To ensure stability, we design MoClip, an optimizer that caps gradient-momentum rotation and applies an arctan2-based rescaling of updates. Empirically, our method improves domain performance on Communication Technology (CT) and NL2SQL tasks while mitigating forgetting on general benchmarks (MMLU, BBH, GPQA, MATH), and reduces training cost by over 40%. Together, these contributions offer a scalable, architecture-agnostic path for domain adaptation of LLMs without sacrificing generalization.

Suming Qiu, Jing Li, Zhicheng Zhou et al.

We present HES-SQL, a novel hybrid training framework that advances Text-to-SQL generation through the integration of thinking-mode-fused supervised fine-tuning (SFT) with Group Relative Policy Optimization (GRPO). Our approach introduces three key innovations: (1) a skeleton-completeness scoring mechanism that enhances preference alignment between generated queries and optimal SQL structures; (2) a query-latency-aware reward system that incentivizes the generation of computationally efficient SQL queries; (3) a self-distillation process for thinking-mode completion that prevents degradation of the model's reasoning capabilities. This framework enables hybrid thinking models to switch between reasoning and non-reasoning modes while improving SQL query accuracy and execution efficiency. Experimental evaluation, conducted on MySQL 8.0 and SQLite 3.42 under controlled single-user conditions, demonstrates that HES-SQL achieves competitive performance with execution accuracies of 79.14\% and 54.9\% on the BIRD and KaggleDBQA benchmarks, respectively. Query latency is measured as the end-to-end execution time of generated queries on the DBMS, averaged over multiple runs to mitigate variance. Efficiency gains range from 11\% to 20\% relative to supervised baselines. Our results establish a new paradigm for Text-to-SQL systems that effectively balances semantic accuracy with computational efficiency through execution-informed reinforcement learning (RL). The proposed methodology has significant implications for developing robust natural language interfaces to databases and can be extended to broader structured generation tasks requiring both correctness and efficiency optimization.

Jing Li, Zhijie Sun, Xuan He et al.

The Mixtures-of-Experts (MoE) model is a widespread distributed and integrated learning method for large language models (LLM), which is favored due to its ability to sparsify and expand models efficiently. However, the performance of MoE is limited by load imbalance and high latency of All-to-All communication, along with relatively redundant computation owing to large expert capacity. Load imbalance may result from existing routing policies that consistently tend to select certain experts. The frequent inter-node communication in the All-to-All procedure also significantly prolongs the training time. To alleviate the above performance problems, we propose a novel routing strategy that combines load balance and locality by converting partial inter-node communication to that of intra-node. Notably, we elucidate that there is a minimum threshold for expert capacity, calculated through the maximal angular deviation between the gating weights of the experts and the assigned tokens. We port these modifications on the PanGu-Sigma model based on the MindSpore framework with multi-level routing and conduct experiments on Ascend clusters. The experiment results demonstrate that the proposed LocMoE reduces training time per epoch by 12.68% to 22.24% compared to classical routers, such as hash router and switch router, without impacting the model accuracy.

Dun Li, Dezhi Han, Noel Crespi et al.

Supply chain finance(SCF) is committed to providing credit for small and medium-sized enterprises(SMEs) with low credit lines and small financing scales. The resulting financial credit data and related business transaction data are highly confidential and private. However, traditional SCF management schemes mostly use third-party platforms and centralized designs, which cannot achieve highly reliable secure storage and fine-grained access control. To fill this gap, this paper designs and implements Fabric-SCF, a secure storage and access control system based on blockchain and attribute-based access control (\textbf{ABAC}) model. This scheme uses distributed consensus to realize data security, traceability, and immutability. We also use smart contracts to define system processes and access policies to ensure the efficient operation of the system. To verify the performance of Fabric-SCF, we designed two sets of simulation experiments. The results show that Fabric-SCF achieves dynamic and fine-grained access control while maintaining high throughput in a simulated real-world operating scenario.