An official website of the United States government
In the context of the rising interest in code language models (code LMs) and vulnerability detection, we study the effectiveness of code LMs for detecting vulnerabilities. Our analysis reveals significant shortcomings in existing vulnerability datasets, including poor data quality, low label accuracy, and high duplication rates, leading to unreliable model performance in realistic vulnerability detection scenarios. Additionally, the evaluation methods used with these datasets are not representative of real-world vulnerability detection. To address these challenges, we introduce PRIMEVUL, a new dataset for training and evaluating code LMs for vulnerability detection. PRIMEVUL incorporates a novel set of data labeling techniques that achieve comparable label accuracy to humanverified benchmarks while significantly expanding the dataset. It also implements a rigorous data de-duplication and chronological data splitting strategy to mitigate data leakage issues, alongside introducing more realistic evaluation metrics and settings. This comprehensive approach aims to provide a more accurate assessment of code LMs’ performance in real-world conditions. Evaluating code LMs on PRIMEVUL reveals that existing benchmarks significantly overestimate the performance of these models. For instance, a state-of-the-art 7B model scored 68.26% F1 on BigVul but only 3.09% F1 on PRIMEVUL. Attempts to improve performance through advanced training techniques and larger models like GPT-3.5 and GPT-4 were unsuccessful, with results akin to random guessing in the most stringent settings. These findings underscore the considerable gap between current capabilities and the practical requirements for deploying code LMs in security roles, highlighting the need for more innovative research in this domain.
more »
« less
Unmasking and Improving Data Credibility: A Study with Datasets for Training Harmless Language Models
Language models have shown promise in various tasks but can be affected by undesired data during training, fine-tuning, or alignment. For example, if some unsafe conversations are wrongly annotated as safe ones, the model fine-tuned on these samples may be harmful. Therefore, the correctness of annotations, i.e., the credibility of the dataset, is important. This study focuses on the credibility of real-world datasets, including the popular benchmarks Jigsaw Civil Comments, Anthropic Harmless & Red Team, PKU BeaverTails & SafeRLHF, that can be used for training a harmless language model. Given the cost and difficulty of cleaning these datasets by humans, we introduce a systematic framework for evaluating the credibility of datasets, identifying label errors, and evaluating the influence of noisy labels in the curated language data, specifically focusing on unsafe comments and conversation classification. With the framework, we find and fix an average of 6.16% label errors in 11 datasets constructed from the above benchmarks. The data credibility and downstream learning performance can be remarkably improved by directly fixing label errors, indicating the significance of cleaning existing real-world datasets.
more »
« less
- PAR ID:
- 10546388
- Publisher / Repository:
- International Conference on Learning Representations
- Date Published:
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
In recent years, deep learning has achieved tremendous success in image segmentation for computer vision applications. The performance of these models heavily relies on the availability of large-scale high-quality training labels (e.g., PASCAL VOC 2012). Unfortunately, such large-scale high-quality training data are often unavailable in many real-world spatial or spatiotemporal problems in earth science and remote sensing (e.g., mapping the nationwide river streams for water resource management). Although extensive efforts have been made to reduce the reliance on labeled data (e.g., semi-supervised or unsupervised learning, few-shot learning), the complex nature of geographic data such as spatial heterogeneity still requires sufficient training labels when transferring a pre-trained model from one region to another. On the other hand, it is often much easier to collect lower-quality training labels with imperfect alignment with earth imagery pixels (e.g., through interpreting coarse imagery by non-expert volunteers). However, directly training a deep neural network on imperfect labels with geometric annotation errors could significantly impact model performance. Existing research that overcomes imperfect training labels either focuses on errors in label class semantics or characterizes label location errors at the pixel level. These methods do not fully incorporate the geometric properties of label location errors in the vector representation. To fill the gap, this article proposes a weakly supervised learning framework to simultaneously update deep learning model parameters and infer hidden true vector label locations. Specifically, we model label location errors in the vector representation to partially reserve geometric properties (e.g., spatial contiguity within line segments). Evaluations on real-world datasets in the National Hydrography Dataset (NHD) refinement application illustrate that the proposed framework outperforms baseline methods in classification accuracy.more » « less
-
Language is compositional; an instruction can ex- press multiple relation constraints to hold among objects in a scene that a robot is tasked to rearrange. Our focus in this work is an instructable scene-rearranging framework that gen- eralizes to longer instructions and to spatial concept compositions never seen at training time. We propose to represent language- instructed spatial concepts with energy functions over relative object arrangements. A language parser maps instructions to corresponding energy functions and an open-vocabulary visual- language model grounds their arguments to relevant objects in the scene. We generate goal scene configurations by gradient descent on the sum of energy functions, one per language predi- cate in the instruction. Local vision-based policies then re-locate objects to the inferred goal locations. We test our model on es- tablished instruction-guided manipulation benchmarks, as well as benchmarks of compositional instructions we introduce. We show our model can execute highly compositional instructions zero-shot in simulation and in the real world. It outperforms language- to-action reactive policies and Large Language Model planners by a large margin, especially for long instructions that involve compositions of multiple spatial concepts. Simulation and real- world robot execution videos, as well as our code and datasets are publicly available on our website: https://ebmplanner.github.io.more » « less
-
Aligning large language models (LLMs) to human preferences is a crucial step in building helpful and safe AI tools, which usually involve training on supervised datasets. Popular algorithms such as Direct Preference Optimization (DPO) rely on pairs of AI-generated responses ranked according to human annotation. The response pair annotation process might bring human bias. Building a correct preference dataset is the costly part of the alignment pipeline. To improve annotation efficiency and quality in the LLMs alignment, we propose REAL:Response Embedding-based Alignment for LLMs, a strategy for constructing a high-quality training dataset that focuses on acquiring the less ambiguous preference pairs for labeling out of a set of response candidates. Our selection process is based on the similarity of embedding responses independently of prompts, which guarantees the selection process in an off-policy setting, avoiding adaptively measuring the similarity during the training. Experimental results on real-world dataset SHP2 and synthetic HH-RLHF benchmarks indicate that choosing dissimilar response pairs enhances the direct alignment of LLMs while reducing inherited labeling errors. The model aligned with dissimilar response pairs obtained a better margin and win rate on the dialogue task. Our findings suggest that focusing on distinct pairs can reduce the label error and improve LLM alignment efficiency, saving up to 65% of annotators’ work. The code of the work can be found https://github.com/ honggen-zhang/REAL-Alignment.more » « less
-
Existing deep-learning approaches to semantic column type annotation (CTA) have important shortcomings: they rely on semantic types which are fixed at training time; require a large number of training samples per type; incur high run-time inference costs; and their performance can degrade when evaluated on novel datasets, even when types remain constant. Large language models have exhibited strong zero-shot classification performance on a wide range of tasks and in this paper we explore their use for CTA. We introduce ArcheType, a simple, practical method for context sampling, prompt serialization, model querying, and label remapping, which enables large language models to solve CTA problems in a fully zero-shot manner. We ablate each component of our method separately, and establish that improvements to context sampling and label remapping provide the most consistent gains. ArcheType establishes a new state-of-the-art performance on zero-shot CTA benchmarks (including three new domain-specific benchmarks which we release along with this paper), and when used in conjunction with classical CTA techniques, it outperforms a SOTA DoDuo model on the fine-tuned SOTAB benchmark.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
