XKV: Personalized KV Cache Memory Reduction for Long-Context LLM Inference

Recently the generative Large Language Model (LLM) has achieved remarkable success in numerous applications. Notably its inference generates output tokens one-by-one, leading to many redundant computations. The widely-used KV-Cache framework makes a compromise between time and space complexities. However, caching data generates the increasingly growing memory demand, that can quickly exhaust the limited memory capacity of the modern accelerator like GPUs, particularly in long-context inference tasks. Existing studies reduce memory consumption by evicting some of cached data that have less important impact on inference accuracy. But the benefit in practice is far from ideal due to the static cache allocation across different LLM network layers. This paper observes that the layer-specific cached data have very different impacts on accuracy. We quantify this difference, and give experimental and theoretical validation. We accordingly make a formal analysis and shows that customizing the cache size for each layer in a personalized manner can yield a significant memory reduction, while still providing comparable accuracy. We simulate the cache allocation as a combinatorial optimization problem and give a global optimal solution. In particular, we devise a mini- and sampling-based inference over a lightweight variant of the LLM model, so as to quickly capture the difference and then feed it into the personalized algorithms. Extensive experiments on real-world datasets demonstrate that our proposals can reduce KV cache memory consumption by 61.6% on average, improve computational efficiency by 2.1x and then increase the throughput by up to 5.5x.

Building A Coding Assistant via the Retrieval-Augmented Language Model

Pretrained language models have shown strong effectiveness in code-related tasks, such as code retrieval, code generation, code summarization, and code completion tasks. In this paper, we propose COde assistaNt viA retrieval-augmeNted language model (CONAN), which aims to build a code assistant by mimicking the knowledge-seeking behaviors of humans during coding. Specifically, it consists of a code structure aware retriever (CONAN-R) and a dual-view code representation-based retrieval-augmented generation model (CONAN-G). CONAN-R pretrains CodeT5 using Code-Documentation Alignment and Masked Entity Prediction tasks to make language models code structure-aware and learn effective representations for code snippets and documentation. Then CONAN-G designs a dual-view code representation mechanism for implementing a retrieval-augmented code generation model. CONAN-G regards the code documentation descriptions as prompts, which help language models better understand the code semantics. Our experiments show that CONAN achieves convincing performance on different code generation tasks and significantly outperforms previous retrieval augmented code generation models. Our further analyses show that CONAN learns tailored representations for both code snippets and documentation by aligning code-documentation data pairs and capturing structural semantics by masking and predicting entities in the code data. Additionally, the retrieved code snippets and documentation provide necessary information from both program language and natural language to assist the code generation process. CONAN can also be used as an assistant for Large Language Models (LLMs), providing LLMs with external knowledge in shorter code document lengths to improve their effectiveness on various code tasks. It shows the ability of CONAN to extract necessary information and help filter out the noise from retrieved code documents.

RAG-DDR: Optimizing Retrieval-Augmented Generation Using Differentiable Data Rewards

Retrieval-Augmented Generation (RAG) has proven its effectiveness in mitigating hallucinations in Large Language Models (LLMs) by retrieving knowledge from external resources. To adapt LLMs for RAG pipelines, current approaches use instruction tuning to optimize LLMs, improving their ability to utilize retrieved knowledge. This supervised fine-tuning (SFT) approach focuses on equipping LLMs to handle diverse RAG tasks using different instructions. However, it trains RAG modules to overfit training signals and overlooks the varying data preferences among agents within the RAG system. In this paper, we propose a Differentiable Data Rewards (DDR) method, which end-to-end trains RAG systems by aligning data preferences between different RAG modules. DDR works by collecting the rewards to optimize each agent with a rollout method. This method prompts agents to sample some potential responses as perturbations, evaluates the impact of these perturbations on the whole RAG system, and subsequently optimizes the agent to produce outputs that improve the performance of the RAG system. Our experiments on various knowledge-intensive tasks demonstrate that DDR significantly outperforms the SFT method, particularly for LLMs with smaller-scale parameters that depend more on the retrieved knowledge. Additionally, DDR exhibits a stronger capability to align the data preference between RAG modules. The DDR method makes generation module more effective in extracting key information from documents and mitigating conflicts between parametric memory and external knowledge. All codes are available at https://github.com/OpenMatch/RAG-DDR.

Enhancing the Code Debugging Ability of LLMs via Communicative Agent Based Data Refinement

Debugging is a vital aspect of software development, yet the debugging capabilities of Large Language Models (LLMs) remain largely unexplored. This paper first introduces DEBUGEVAL, a comprehensive benchmark designed to evaluate the debugging capabilities of LLMs. DEBUGEVAL collects data from existing high-quality datasets and designs four different tasks to evaluate the debugging effectiveness, including BUG Localization, BUG Identification, Code Review, and Code Repair. Additionally, to enhance the code debugging ability of LLMs, this paper proposes a CoMmunicative Agent BaSed DaTa REfinement FRamework (MASTER), which generates the refined code debugging data for supervised finetuning. Specifically, MASTER employs the Code Quizzer to generate refined data according to the defined tasks of DEBUGEVAL. Then the Code Learner acts as a critic and reserves the generated problems that it can not solve. Finally, the Code Teacher provides a detailed Chain-of-Thought based solution to deal with the generated problem. We collect the synthesized data and finetune the Code Learner to enhance the debugging ability and conduct the NeuDebugger model. Our experiments evaluate various LLMs and NeuDebugger in the zero-shot setting on DEBUGEVAL. Experimental results demonstrate that these 7B-scale LLMs have weaker debugging capabilities, even these code-oriented LLMs. On the contrary, these larger models (over 70B) show convincing debugging ability. Our further analyses illustrate that MASTER is an effective method to enhance the code debugging ability by synthesizing data for Supervised Fine-Tuning (SFT) LLMs.

Affective Computing in the Era of Large Language Models: A Survey from the NLP Perspective

Affective Computing (AC), integrating computer science, psychology, and cognitive science knowledge, aims to enable machines to recognize, interpret, and simulate human emotions.To create more value, AC can be applied to diverse scenarios, including social media, finance, healthcare, education, etc. Affective Computing (AC) includes two mainstream tasks, i.e., Affective Understanding (AU) and Affective Generation (AG). Fine-tuning Pre-trained Language Models (PLMs) for AU tasks has succeeded considerably. However, these models lack generalization ability, requiring specialized models for specific tasks. Additionally, traditional PLMs face challenges in AG, particularly in generating diverse and emotionally rich responses. The emergence of Large Language Models (LLMs), such as the ChatGPT series and LLaMA models, brings new opportunities and challenges, catalyzing a paradigm shift in AC. LLMs possess capabilities of in-context learning, common sense reasoning, and advanced sequence generation, which present unprecedented opportunities for AU. To provide a comprehensive overview of AC in the LLMs era from an NLP perspective, we summarize the development of LLMs research in this field, aiming to offer new insights. Specifically, we first summarize the traditional tasks related to AC and introduce the preliminary study based on LLMs. Subsequently, we outline the relevant techniques of popular LLMs to improve AC tasks, including Instruction Tuning and Prompt Engineering. For Instruction Tuning, we discuss full parameter fine-tuning and parameter-efficient methods such as LoRA, P-Tuning, and Prompt Tuning. In Prompt Engineering, we examine Zero-shot, Few-shot, Chain of Thought (CoT), and Agent-based methods for AU and AG. To clearly understand the performance of LLMs on different Affective Computing tasks, we further summarize the existing benchmarks and evaluation methods.

Multi-Evidence based Fact Verification via A Confidential Graph Neural Network

Fact verification tasks aim to identify the integrity of textual contents according to the truthful corpus. Existing fact verification models usually build a fully connected reasoning graph, which regards claim-evidence pairs as nodes and connects them with edges. They employ the graph to propagate the semantics of the nodes. Nevertheless, the noisy nodes usually propagate their semantics via the edges of the reasoning graph, which misleads the semantic representations of other nodes and amplifies the noise signals. To mitigate the propagation of noisy semantic information, we introduce a Confidential Graph Attention Network (CO-GAT), which proposes a node masking mechanism for modeling the nodes. Specifically, CO-GAT calculates the node confidence score by estimating the relevance between the claim and evidence pieces. Then, the node masking mechanism uses the node confidence scores to control the noise information flow from the vanilla node to the other graph nodes. CO-GAT achieves a 73.59% FEVER score on the FEVER dataset and shows the generalization ability by broadening the effectiveness to the science-specific domain.

Say More with Less: Understanding Prompt Learning Behaviors through Gist Compression

Large language models (LLMs) require lengthy prompts as the input context to produce output aligned with user intentions, a process that incurs extra costs during inference. In this paper, we propose the Gist COnditioned deCOding (Gist-COCO) model, introducing a novel method for compressing prompts which also can assist the prompt interpretation and engineering. Gist-COCO employs an encoder-decoder based language model and then incorporates an additional encoder as a plugin module to compress prompts with inputs using gist tokens. It finetunes the compression plugin module and uses the representations of gist tokens to emulate the raw prompts in the vanilla language model. By verbalizing the representations of gist tokens into gist prompts, the compression ability of Gist-COCO can be generalized to different LLMs with high compression rates. Our experiments demonstrate that Gist-COCO outperforms previous prompt compression models in both passage and instruction compression tasks. Further analysis on gist verbalization results suggests that our gist prompts serve different functions in aiding language models. They may directly provide potential answers, generate the chain-of-thought, or simply repeat the inputs. All data and codes are available at https://github.com/OpenMatch/Gist-COCO .

Cleaner Pretraining Corpus Curation with Neural Web Scraping

The web contains large-scale, diverse, and abundant information to satisfy the information-seeking needs of humans. Through meticulous data collection, preprocessing, and curation, webpages can be used as a fundamental data resource for language model pretraining. However, when confronted with the progressively revolutionized and intricate nature of webpages, rule-based/feature-based web scrapers are becoming increasingly inadequate. This paper presents a simple, fast, and effective Neural web Scraper (NeuScraper) to help extract primary and clean text contents from webpages. Experimental results show that NeuScraper surpasses the baseline scrapers by achieving more than a 20% improvement, demonstrating its potential in extracting higher-quality data to facilitate the language model pretraining. All of the code is available at https://github.com/OpenMatch/NeuScraper.

ActiveRAG: Revealing the Treasures of Knowledge via Active Learning

Retrieval Augmented Generation (RAG) has introduced a new paradigm for Large Language Models (LLMs), aiding in the resolution of knowledge-intensive tasks. However, current RAG models position LLMs as passive knowledge receptors, thereby restricting their capacity for learning and comprehending external knowledge. In this paper, we present ActiveRAG, an innovative RAG framework that shifts from passive knowledge acquisition to an active learning mechanism. This approach utilizes the Knowledge Construction mechanism to develop a deeper understanding of external knowledge by associating it with previously acquired or memorized knowledge. Subsequently, it designs the Cognitive Nexus mechanism to incorporate the outcomes from both chains of thought and knowledge construction, thereby calibrating the intrinsic cognition of LLMs. Our experimental results demonstrate that ActiveRAG surpasses previous RAG models, achieving a 5% improvement on question-answering datasets. All data and codes are available at https://github.com/OpenMatch/ActiveRAG.

LegalDuet: Learning Effective Representations for Legal Judgment Prediction through a Dual-View Legal Clue Reasoning

Most existing Legal Judgment Prediction (LJP) models focus on discovering the legal triggers in the criminal fact description. However, in real-world scenarios, a professional judge not only needs to assimilate the law case experience that thrives on past sentenced legal judgments but also depends on the professional legal grounded reasoning that learned from professional legal knowledge. In this paper, we propose a LegalDuet model, which pretrains language models to learn a tailored embedding space for making legal judgments. It proposes a dual-view legal clue reasoning mechanism, which derives from two reasoning chains of judges: 1) Law Case Reasoning, which makes legal judgments according to the judgment experiences learned from analogy/confusing legal cases; 2) Legal Ground Reasoning, which lies in matching the legal clues between criminal cases and legal decisions. Our experiments show that LegalDuet achieves state-of-the-art performance on the CAIL2018 dataset and outperforms baselines with about 4% improvements on average. Our dual-view reasoning based pretraining can capture critical legal clues to learn a tailored embedding space to distinguish criminal cases. It reduces LegalDuet's uncertainty during prediction and brings pretraining advances to the confusing/low frequent charges. All codes are available at https://github.com/NEUIR/LegalDuet.