PRMBench: A Fine-grained and Challenging Benchmark for Process-Level Reward Models

Process-level Reward Models (PRMs) are crucial for complex reasoning and decision-making tasks, where each intermediate step plays an important role in the reasoning process. Since language models are prone to various types of errors during the reasoning process, PRMs are required to possess nuanced capabilities for detecting various implicit error types in real-world scenarios. However, current benchmarks primarily focus on step correctness, failing to evaluate PRMs' performance systematically. To address this gap, we introduce PRMBench, a process-level benchmark specifically designed to assess the fine-grained error detection capabilities of PRMs. PRMBench comprises 6,216 carefully designed problems and 83,456 step-level labels, evaluating models across multiple dimensions, including simplicity, soundness, and sensitivity. In our experiments on 15 models, spanning both open-source PRMs and closed-source large language models prompted as critic models, we uncover significant weaknesses in current PRMs. These findings underscore the challenges inherent in process-level evaluation and highlight key directions for future research. We hope PRMBench can be a robust bench for advancing research on PRM evaluation and development.

A Survey of Query Optimization in Large Language Models

\textit{Query Optimization} (QO) refers to techniques aimed at enhancing the efficiency and quality of Large Language Models (LLMs) in understanding and answering queries, especially complex ones in scenarios like Retrieval-Augmented Generation (RAG). Specifically, RAG mitigates the limitations of LLMs by dynamically retrieving and leveraging up-to-date relevant information, which provides a cost-effective solution to the challenge of LLMs producing plausible but potentially inaccurate responses. Recently, as RAG evolves and incorporates multiple components that influence its performance, QO has emerged as a critical element, playing a pivotal role in determining the effectiveness of RAG's retrieval stage in accurately sourcing the necessary multiple pieces of evidence to answer queries correctly. In this paper, we trace the evolution of QO techniques by summarizing and analyzing significant studies. Through an organized framework and categorization, we aim to consolidate existing QO techniques in RAG, elucidate their technological foundations, and highlight their potential to enhance the versatility and applications of LLMs.

MiMoTable: A Multi-scale Spreadsheet Benchmark with Meta Operations for Table Reasoning

Extensive research has been conducted to explore the capability of Large Language Models (LLMs) for table reasoning and has significantly improved the performance on existing benchmarks. However, tables and user questions in real-world applications are more complex and diverse, presenting an unignorable gap compared to the existing benchmarks. To fill the gap, we propose a \textbf{M}ult\textbf{i}-scale spreadsheet benchmark with \textbf{M}eta \textbf{o}perations for \textbf{Table} reasoning, named as MiMoTable. Specifically, MiMoTable incorporates two key features. First, the tables in MiMoTable are all spreadsheets used in real-world scenarios, which cover seven domains and contain different types. Second, we define a new criterion with six categories of meta operations for measuring the difficulty of each question in MiMoTable, simultaneously as a new perspective for measuring the difficulty of the existing benchmarks. Experimental results show that Claude-3.5-Sonnet achieves the best performance with 77.4\% accuracy, indicating that there is still significant room to improve for LLMs on MiMoTable. Furthermore, we grade the difficulty of existing benchmarks according to our new criteria. Experiments have shown that the performance of LLMs decreases as the difficulty of benchmarks increases, thereby proving the effectiveness of our proposed new criterion.

P/D-Serve: Serving Disaggregated Large Language Model at Scale

Serving disaggregated large language models (LLMs) over tens of thousands of xPU devices (GPUs or NPUs) with reliable performance faces multiple challenges. 1) Ignoring the diversity (various prefixes and tidal requests), treating all the prompts in a mixed pool is inadequate. To facilitate the similarity per scenario and minimize the inner mismatch on P/D (prefill and decoding) processing, fine-grained organization is required, dynamically adjusting P/D ratios for better performance. 2) Due to inaccurate estimation on workload (queue status or maintained connections), the global scheduler easily incurs unnecessary timeouts in prefill. 3) Block-fixed device-to-device (D2D) KVCache transfer over cluster-level RDMA (remote direct memory access) fails to achieve desired D2D utilization as expected. To overcome previous problems, this paper proposes an end-to-end system P/D-Serve, complying with the paradigm of MLOps (machine learning operations), which models end-to-end (E2E) P/D performance and enables: 1) fine-grained P/D organization, mapping the service with RoCE (RDMA over converged ethernet) as needed, to facilitate similar processing and dynamic adjustments on P/D ratios; 2) on-demand forwarding upon rejections for idle prefill, decoupling the scheduler from regular inaccurate reports and local queues, to avoid timeouts in prefill; and 3) efficient KVCache transfer via optimized D2D access. P/D-Serve is implemented upon Ascend and MindSpore, has been deployed over tens of thousands of NPUs for more than eight months in commercial use, and further achieves 60\%, 42\% and 46\% improvements on E2E throughput, time-to-first-token (TTFT) SLO (service level objective) and D2D transfer time. As the E2E system with optimizations, P/D-Serve achieves 6.7x increase on throughput, compared with aggregated LLMs.

Mitigating Multilingual Hallucination in Large Vision-Language Models

While Large Vision-Language Models (LVLMs) have exhibited remarkable capabilities across a wide range of tasks, they suffer from hallucination problems, where models generate plausible yet incorrect answers given the input image-query pair. This hallucination phenomenon is even more severe when querying the image in non-English languages, while existing methods for mitigating hallucinations in LVLMs only consider the English scenarios. In this paper, we make the first attempt to mitigate this important multilingual hallucination in LVLMs. With thorough experiment analysis, we found that multilingual hallucination in LVLMs is a systemic problem that could arise from deficiencies in multilingual capabilities or inadequate multimodal abilities. To this end, we propose a two-stage Multilingual Hallucination Removal (MHR) framework for LVLMs, aiming to improve resistance to hallucination for both high-resource and low-resource languages. Instead of relying on the intricate manual annotations of multilingual resources, we fully leverage the inherent capabilities of the LVLM and propose a novel cross-lingual alignment method, which generates multiple responses for each image-query input and then identifies the hallucination-aware pairs for each language. These data pairs are finally used for direct preference optimization to prompt the LVLMs to favor non-hallucinating responses. Experimental results show that our MHR achieves a substantial reduction in hallucination generation for LVLMs. Notably, on our extended multilingual POPE benchmark, our framework delivers an average increase of 19.0% in accuracy across 13 different languages. Our code and model weights are available at https://github.com/ssmisya/MHR

SS-Bench: A Benchmark for Social Story Generation and Evaluation

Children with Autism Spectrum Disorder (ASD) often misunderstand social situations and struggle to participate in daily routines. Psychology experts write Social Stories under strict constraints of structural clarity, descriptive orientation, and situational safety to enhance their abilities in these regimes. However, Social Stories are costly in creation and often limited in diversity and timeliness. As Large Language Models (LLMs) become increasingly powerful, there is a growing need for more automated, affordable, and accessible methods to generate Social Stories in real-time with broad coverage. Adapting LLMs to meet the unique and strict constraints of Social Stories is a challenging issue. To this end, we propose \textbf{SS-Bench}, a \textbf{S}ocial \textbf{S}tory \textbf{Bench}mark for generating and evaluating Social Stories. Specifically, we develop a constraint-driven strategy named \textbf{\textsc{StarSow}} to hierarchically prompt LLMs to generate Social Stories and build a benchmark, which has been validated through experiments to fine-tune smaller models for generating qualified Social Stories. Additionally, we introduce \textbf{Quality Assessment Criteria}, employed in human and GPT evaluations, to verify the effectiveness of the generated stories. We hope this work benefits the autism community and catalyzes future research focusing on particular groups.

Can Many-Shot In-Context Learning Help Long-Context LLM Judges? See More, Judge Better!

Leveraging Large Language Models (LLMs) as judges for evaluating the performance of LLMs has recently garnered attention. Nonetheless, this type of approach concurrently introduces potential biases from LLMs, raising concerns about the reliability of the evaluation results. To mitigate this issue, we propose and study two versions of many-shot in-context prompts, Reinforced and Unsupervised ICL, for helping GPT-4o-as-a-Judge in single answer grading. Based on the designed prompts, we investigate the impact of scaling the number of in-context examples on the agreement and quality of the evaluation. Furthermore, we first reveal the symbol bias in GPT-4o-as-a-Judge for pairwise comparison and then propose a simple yet effective approach to mitigate it. Experimental results show that advanced long-context LLMs, such as GPT-4o, perform better in the many-shot regime than in the zero-shot regime. Meanwhile, the experimental results further verify the effectiveness of the symbol bias mitigation approach.

A Preliminary Empirical Study on Prompt-based Unsupervised Keyphrase Extraction

Pre-trained large language models can perform natural language processing downstream tasks by conditioning on human-designed prompts. However, a prompt-based approach often requires "prompt engineering" to design different prompts, primarily hand-crafted through laborious trial and error, requiring human intervention and expertise. It is a challenging problem when constructing a prompt-based keyphrase extraction method. Therefore, we investigate and study the effectiveness of different prompts on the keyphrase extraction task to verify the impact of the cherry-picked prompts on the performance of extracting keyphrases. Extensive experimental results on six benchmark keyphrase extraction datasets and different pre-trained large language models demonstrate that (1) designing complex prompts may not necessarily be more effective than designing simple prompts; (2) individual keyword changes in the designed prompts can affect the overall performance; (3) designing complex prompts achieve better performance than designing simple prompts when facing long documents.

Counting-Stars: A Simple, Efficient, and Reasonable Strategy for Evaluating Long-Context Large Language Models

While recent research endeavors have concentrated on developing Large Language Models (LLMs) with robust long-context capabilities, due to the lack of appropriate evaluation strategies, relatively little is known about how well the long-context capability and performance of leading LLMs (e.g., GPT-4 Turbo and Kimi Chat). To address this gap, we propose a simple, efficient, and reasonable strategy for evaluating long-context LLMs as a new benchmark, named Counting-Stars. The Counting-Stars is designed to require LLMs to fully understand and capture long dependencies in long contexts, further being able to collect inter-dependency across multiple pieces of evidence spanning the entire context to finish the task. Based on the Counting-Stars, we conduct experiments to evaluate the two leading long-context LLMs, i.e., GPT-4 Turbo and Kimi Chat. The experimental results indicate that GPT-4 Turbo and Kimi Chat achieve significant performance in the long context from 4K to 128K. We further present several intriguing analyses regarding the behavior of LLMs processing long context.

Large Language Models as Zero-Shot Keyphrase Extractors: A Preliminary Empirical Study

Zero-shot keyphrase extraction aims to build a keyphrase extractor without training by human-annotated data, which is challenging due to the limited human intervention involved. Challenging but worthwhile, zero-shot setting efficiently reduces the time and effort that data labeling takes. Recent efforts on pre-trained large language models (e.g., ChatGPT and ChatGLM) show promising performance on zero-shot settings, thus inspiring us to explore prompt-based methods. In this paper, we ask whether strong keyphrase extraction models can be constructed by directly prompting the large language model ChatGPT. Through experimental results, it is found that ChatGPT still has a lot of room for improvement in the keyphrase extraction task compared to existing state-of-the-art unsupervised and supervised models.