Motivated by the increasing popularity of code generation from human descriptions using large language models (LLMs), several benchmarks have been proposed to assess the capabilities of existing and emerging models. This study presents a large-scale human evaluation of HumanEval and MBPP, two widely used benchmarks for Python code generation, focusing on their diversity and difficulty. Our findings reveal a significant bias towards a limited number of programming concepts, with negligible or no representation of most concepts. Additionally, we identify a concerningly high proportion of easy programming questions, potentially leading to an overestimation of model performance on code generation tasks.
相關內容
Large language models (LLMs) have exhibited remarkable ability in code generation. However, generating the correct solution in a single attempt still remains a challenge. Prior works utilize verification properties in software engineering to verify and re-rank solutions in a majority voting manner. But the assumption behind them that generated verification properties have better qualities than solutions may not always hold. In this paper, we treat them equally as different perspectives of LLMs' reasoning processes. We propose the Multi-Perspective Self-Consistency (MPSC) framework incorporating both inter- and intra-consistency across outputs from multiple perspectives. Specifically, we prompt LLMs to generate diverse outputs from three perspectives, Solution, Specification and Test case, constructing a 3-partite graph. With two measure functions of consistency, we embed both inter- and intra-consistency information into the graph. The optimal choice of solutions is then determined based on analysis in the graph. MPSC significantly boosts performance of foundation models (ChatGPT in this paper) on various benchmarks, including HumanEval (+15.91%), MBPP (+6.43%) and CodeContests (+9.37%), even surpassing GPT-4.
Recently, numerous new benchmarks have been established to evaluate the performance of large language models (LLMs) via either computing a holistic score or employing another LLM as a judge. However, these approaches suffer from data leakage due to the open access of the benchmark and inflexible evaluation process. To address this issue, we introduce $\textbf{TreeEval}$, a benchmark-free evaluation method for LLMs that let a high-performance LLM host an irreproducible evaluation session and essentially avoids the data leakage. Moreover, this LLM performs as an examiner to raise up a series of questions under a topic with a tree planing strategy, which considers the current evaluation status to decide the next question generation and ensures the completeness and efficiency of the evaluation process. We evaluate $6$ models of different parameter sizes, including $7$B, $13$B, and $33$B, and ultimately achieved the highest correlation coefficient with AlpacaEval2.0 using only around $45$ questions. We also conduct more analysis to show the robustness and reliability of TreeEval. Our code can be accessed via the provided //github.com/Ashura5/TreeEval.
Large language models (LLMs) have made significant strides in reasoning capabilities, with ongoing efforts to refine their reasoning through self-correction. However, recent studies suggest that self-correction can be limited or even counterproductive without external accurate knowledge, raising questions about the limits and effectiveness of self-correction. In this paper, we aim to enhance LLM's self-checking capabilities by meticulously designing training data, thereby improving the accuracy of self-correction. We conduct a detailed analysis of error types in mathematical reasoning and develop a tailored prompt, termed ``Step CoT Check''. Then we construct a checking-correction dataset for training models. After integrating the original CoT data and checking-correction data for training, we observe that models could improve their self-checking capabilities, thereby enhancing their self-correction capacity and eliminating the need for external feedback or ground truth labels to ascertain the endpoint of correction. We compare the performance of models fine-tuned with the ``Step CoT Check'' prompt against those refined using other promps within the context of checking-correction data. The ``Step CoT Check'' outperforms the other two check formats in model with lager parameters, providing more precise feedback thus achieving a higher rate of correctness. For reproducibility, all the datasets and codes are provided in \url{//github.com/bammt/Learn-to-check}.
We analyze the behaviors of open large language models (LLMs) on the task of data-to-text (D2T) generation, i.e., generating coherent and relevant text from structured data. To avoid the issue of LLM training data contamination with standard benchmarks, we design Quintd - a tool for collecting novel structured data records from public APIs. Using a dataset collected with Quintd and leveraging reference-free evaluation, we analyze model behaviors on five D2T generation tasks. We find that recent open LLMs (Llama2, Mistral, and Zephyr) can generate fluent and coherent text from standard data formats in zero-shot settings. However, we also show that the semantic accuracy of the outputs is a major issue: both according to our GPT-4-based metric and human annotators, more than 80% of the outputs of open LLMs contain a semantic error. We publicly release the code, data, and model outputs.
The integration of large language models (LLMs) and search engines represents a significant evolution in knowledge acquisition methodologies. However, determining the knowledge that an LLM already possesses and the knowledge that requires the help of a search engine remains an unresolved issue. Most existing methods solve this problem through the results of preliminary answers or reasoning done by the LLM itself, but this incurs excessively high computational costs. This paper introduces a novel collaborative approach, namely SlimPLM, that detects missing knowledge in LLMs with a slim proxy model, to enhance the LLM's knowledge acquisition process. We employ a proxy model which has far fewer parameters, and take its answers as heuristic answers. Heuristic answers are then utilized to predict the knowledge required to answer the user question, as well as the known and unknown knowledge within the LLM. We only conduct retrieval for the missing knowledge in questions that the LLM does not know. Extensive experimental results on five datasets with two LLMs demonstrate a notable improvement in the end-to-end performance of LLMs in question-answering tasks, achieving or surpassing current state-of-the-art models with lower LLM inference costs.
Catastrophic forgetting emerges as a critical challenge when fine-tuning multi-modal large language models (MLLMs), where improving performance on unseen tasks often leads to a significant performance drop on the original tasks. This paper presents a comprehensive analysis of catastrophic forgetting in MLLMs and introduces a post-training adjustment method called Model Tailor. Our method primarily preserves the pre-trained parameters while replacing a small number ($\leq$ 10\%) of fine-tuned parameters, maintaining $\sim$ 99\% effectiveness on original tasks versus pre-training, and achieving $\sim$ 97\% on new tasks compared to standard fine-tuning. Specifically, we derive a sparse mask to identify the "model patch", based on a fusion strategy that integrates salience and sensitivity analysis. Subsequently, a compensation mechanism is introduced to "decorate the patch", enhancing the model's performance on both target and original tasks. Additionally, our method is adaptable to multi-task scenarios. Through extensive experiments on InstructBLIP and LLaVA-1.5 in both image captioning and visual question answering tasks, our approach demonstrates significant task adaptability while preserving inherent pre-trained capabilities.
We introduce and test the System Dynamics Bot, a computer program leveraging a large language model to automate the creation of causal loop diagrams from textual data. To evaluate its performance, we ensembled two distinct databases. The first dataset includes 20 causal loop diagrams and associated texts sourced from the system dynamics literature. The second dataset comprises responses from 30 participants to a vignette, along with causal loop diagrams coded by three system dynamics modelers. The bot uses textual data and successfully identifies approximately sixty percent of the links between variables and feedback loops in both datasets. This paper outlines our approach, provides examples, and presents evaluation results. We discuss encountered challenges and implemented solutions in developing the System Dynamics Bot. The bot can facilitate extracting mental models from textual data and improve model building processes. Moreover, the two datasets can serve as a testbed for similar programs.
We introduce MultiMedEval, an open-source toolkit for fair and reproducible evaluation of large, medical vision-language models (VLM). MultiMedEval comprehensively assesses the models' performance on a broad array of six multi-modal tasks, conducted over 23 datasets, and spanning over 11 medical domains. The chosen tasks and performance metrics are based on their widespread adoption in the community and their diversity, ensuring a thorough evaluation of the model's overall generalizability. We open-source a Python toolkit (github.com/corentin-ryr/MultiMedEval) with a simple interface and setup process, enabling the evaluation of any VLM in just a few lines of code. Our goal is to simplify the intricate landscape of VLM evaluation, thus promoting fair and uniform benchmarking of future models.
With the advent of large language models (LLM), the line between human-crafted and machine-generated texts has become increasingly blurred. This paper delves into the inquiry of identifying discernible and unique linguistic properties in texts that were written by humans, particularly uncovering the underlying discourse structures of texts beyond their surface structures. Introducing a novel methodology, we leverage hierarchical parse trees and recursive hypergraphs to unveil distinctive discourse patterns in texts produced by both LLMs and humans. Empirical findings demonstrate that, although both LLMs and humans generate distinct discourse patterns influenced by specific domains, human-written texts exhibit more structural variability, reflecting the nuanced nature of human writing in different domains. Notably, incorporating hierarchical discourse features enhances binary classifiers' overall performance in distinguishing between human-written and machine-generated texts, even on out-of-distribution and paraphrased samples. This underscores the significance of incorporating hierarchical discourse features in the analysis of text patterns. The code and dataset will be available at [TBA].
The emergence of large language models (LLMs) has substantially influenced natural language processing, demonstrating exceptional results across various tasks. In this study, we employ ``Introspective Tips" to facilitate LLMs in self-optimizing their decision-making. By introspectively examining trajectories, LLM refines its policy by generating succinct and valuable tips. Our method enhances the agent's performance in both few-shot and zero-shot learning situations by considering three essential scenarios: learning from the agent's past experiences, integrating expert demonstrations, and generalizing across diverse games. Importantly, we accomplish these improvements without fine-tuning the LLM parameters; rather, we adjust the prompt to generalize insights from the three aforementioned situations. Our framework not only supports but also emphasizes the advantage of employing LLM in in-contxt decision-making. Experiments involving over 100 games in TextWorld illustrate the superior performance of our approach.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191