Jianhua Gao

Danying Ge, Jianhua Gao, Yixue Yang et al.

Retrieval-Augmented Generation (RAG) improves model output accuracy by leveraging external knowledge bases, serving as an effective solution to address hallucination issues and knowledge-update delays in Large Language Models (LLMs). However, the introduction of external knowledge bases presents RAG with challenges in long-context processing, significantly increasing memory consumption and inference latency. Existing research accelerates inference by precomputing Key and Value (KV) of the knowledge base and loading them on-demand during inference. Based on the access frequency of different KV chunks within the external knowledge base, this paper proposes a hotness-aware RAG (HA-RAG) inference optimization system. First, leveraging the numerical distribution of KV chunks, we introduce a hotness-aware mixed-precision compressing and loading method to reduce disk I/O and memory access overhead. Second, we design a hotness-aware data placement strategy that prioritizes storing frequently accessed KV chunks in high-speed memory to improve data access efficiency. Experimental results demonstrate that, compared with TurboRAG, the proposed HA-RAG achieves an average speedup of 2.10x and maximum speedup of 10.49x in Time-To-First-Token (TTFT) with negligible accuracy loss.

Danying Ge, Jianhua Gao, Qizhi Jiang et al.

Speculative decoding, which combines a draft model with a target model, has emerged as an effective approach to accelerate large language model (LLM) inference. However, existing methods often face a trade-off between the acceptance rate and decoding speed in downstream tasks due to the limited capacity of the draft model, making it difficult to ensure efficiency across diverse tasks. To address this problem, we propose a speculative decoding algorithm tailored for downstream task optimization. It includes an automatic task partitioning and assigning method, which automatically categorizes downstream tasks into different sub-tasks and assigns them to a set of heterogeneous draft models. Each draft model is aligned with the target model using task-specific data, thereby enhancing the consistency of inference results. In addition, our proposed method incorporates an online lightweight prompt classifier to dynamically route prompts to the appropriate draft model. Experimental results demonstrate that the proposed method improves draft accuracy by 6% to 50% over vanilla speculative decoding, while achieving a speedup of 1.10x to 2.64x in LLM inference.

Zijie Huang, Zhiqing Shao, Guisheng Fan et al.

Community smells appear in sub-optimal software development community structures, causing unforeseen additional project costs, e.g., lower productivity and more technical debt. Previous studies analyzed and predicted community smells in the granularity of community sub-groups using socio-technical factors. However, refactoring such smells requires the effort of developers individually. To eliminate them, supportive measures for every developer should be constructed according to their motifs and working states. Recent work revealed developers' personalities could influence community smells' variation, and their sentiments could impact productivity. Thus, sentiments could be evaluated to predict community smells' occurrence on them. To this aim, this paper builds a developer-oriented and sentiment-aware community smell prediction model considering 3 smells such as Organizational Silo, Lone Wolf, and Bottleneck. Furthermore, it also predicts if a developer quitted the community after being affected by any smell. The proposed model achieves cross- and within-project prediction F-Measure ranging from 76% to 93%. Research also reveals 6 sentimental features having stronger predictive power compared with activeness metrics. Imperative and indicative expressions, politeness, and several emotions are the most powerful predictors. Finally, we test statistically the mean and distribution of sentimental features. Based on our findings, we suggest developers should communicate in a straightforward and polite way.