Finding Safety Neurons in Large Language Models

Large language models (LLMs) excel in various capabilities but also pose safety risks such as generating harmful content and misinformation, even after safety alignment. In this paper, we explore the inner mechanisms of safety alignment from the perspective of mechanistic interpretability, focusing on identifying and analyzing safety neurons within LLMs that are responsible for safety behaviors. We propose generation-time activation contrasting to locate these neurons and dynamic activation patching to evaluate their causal effects. Experiments on multiple recent LLMs show that: (1) Safety neurons are sparse and effective. We can restore $90$% safety performance with intervention only on about $5$% of all the neurons. (2) Safety neurons encode transferrable mechanisms. They exhibit consistent effectiveness on different red-teaming datasets. The finding of safety neurons also interprets "alignment tax". We observe that the identified key neurons for safety and helpfulness significantly overlap, but they require different activation patterns of the shared neurons. Furthermore, we demonstrate an application of safety neurons in detecting unsafe outputs before generation. Our findings may promote further research on understanding LLM alignment. The source codes will be publicly released to facilitate future research.

ChatGLM: A Family of Large Language Models from GLM-130B to GLM-4 All Tools

We introduce ChatGLM, an evolving family of large language models that we have been developing over time. This report primarily focuses on the GLM-4 language series, which includes GLM-4, GLM-4-Air, and GLM-4-9B. They represent our most capable models that are trained with all the insights and lessons gained from the preceding three generations of ChatGLM. To date, the GLM-4 models are pre-trained on ten trillions of tokens mostly in Chinese and English, along with a small set of corpus from 24 languages, and aligned primarily for Chinese and English usage. The high-quality alignment is achieved via a multi-stage post-training process, which involves supervised fine-tuning and learning from human feedback. Evaluations show that GLM-4 1) closely rivals or outperforms GPT-4 in terms of general metrics such as MMLU, GSM8K, MATH, BBH, GPQA, and HumanEval, 2) gets close to GPT-4-Turbo in instruction following as measured by IFEval, 3) matches GPT-4 Turbo (128K) and Claude 3 for long context tasks, and 4) outperforms GPT-4 in Chinese alignments as measured by AlignBench. The GLM-4 All Tools model is further aligned to understand user intent and autonomously decide when and which tool(s) touse -- including web browser, Python interpreter, text-to-image model, and user-defined functions -- to effectively complete complex tasks. In practical applications, it matches and even surpasses GPT-4 All Tools in tasks like accessing online information via web browsing and solving math problems using Python interpreter. Over the course, we have open-sourced a series of models, including ChatGLM-6B (three generations), GLM-4-9B (128K, 1M), GLM-4V-9B, WebGLM, and CodeGeeX, attracting over 10 million downloads on Hugging face in the year 2023 alone. The open models can be accessed through https://github.com/THUDM and https://huggingface.co/THUDM.

DICE: Detecting In-distribution Contamination in LLM's Fine-tuning Phase for Math Reasoning

The advancement of large language models (LLMs) relies on evaluation using public benchmarks, but data contamination can lead to overestimated performance. Previous researches focus on detecting contamination by determining whether the model has seen the exact same data during training. In this work, we argue that even training on data similar to benchmark data inflates performance on in-distribution tasks without improving overall capacity, which we called In-distribution contamination. To effectively detect in-distribution contamination, we propose DICE, a novel method that leverages the internal states of LLMs to locate-then-detect the contamination. DICE first identifies the most sensitive layer to contamination, then trains a classifier based on the internal states of that layer. Experiments reveal DICE's high accuracy in detecting in-distribution contamination across various LLMs and math reasoning datasets. We also show the generalization capability of the trained DICE detector, which is able to detect contamination across multiple benchmarks with similar distributions. Additionally, we find that the DICE detection scores are positively correlated with the performance of ten LLMs fine-tuned by either us or other organizations on four math reasoning datasets (with $R^2$ values between 0.6 and 0.75). This indicates that the in-distribution contamination problem potentially lead to an overestimation of the true capabilities of many existing models. The code and data are available at https://github.com/THU-KEG/DICE.

Advancing Geometric Problem Solving: A Comprehensive Benchmark for Multimodal Model Evaluation

In this work, we present the MM-MATH dataset, a novel benchmark developed to rigorously evaluate the performance of advanced large language and multimodal models - including but not limited to GPT-4, GPT-4V, and Claude - within the domain of geometric computation. This dataset comprises 5,929 meticulously crafted geometric problems, each paired with a corresponding image, aimed at mirroring the complexity and requirements typical of ninth-grade mathematics. The motivation behind MM-MATH stems from the burgeoning interest and significant strides in multimodal technology, which necessitates a paradigm shift in assessment methodologies from mere outcome analysis to a more holistic evaluation encompassing reasoning and procedural correctness. Despite impressive gains in various benchmark performances, our analysis uncovers a persistent and notable deficiency in these models' ability to parse and interpret geometric information accurately from images, accounting for over 60% of observed errors. By deploying a dual-focused evaluation approach, examining both the end results and the underlying problem-solving processes, we unearthed a marked discrepancy between the capabilities of current multimodal models and human-level proficiency. The introduction of MM-MATH represents a tripartite contribution to the field: it not only serves as a comprehensive and challenging benchmark for assessing geometric problem-solving prowess but also illuminates critical gaps in textual and visual comprehension that current models exhibit. Through this endeavor, we aspire to catalyze further research and development aimed at bridging these gaps, thereby advancing the state of multimodal model capabilities to new heights.

CogCoM: Train Large Vision-Language Models Diving into Details through Chain of Manipulations

Vision-Language Models (VLMs) have demonstrated their widespread viability thanks to extensive training in aligning visual instructions to answers. However, this conclusive alignment leads models to ignore critical visual reasoning, and further result in failures on meticulous visual problems and unfaithful responses. In this paper, we propose Chain of Manipulations, a mechanism that enables VLMs to solve problems with a series of manipulations, where each manipulation refers to an operation on the visual input, either from intrinsic abilities (e.g., grounding) acquired through prior training or from imitating human-like behaviors (e.g., zoom in). This mechanism encourages VLMs to generate faithful responses with evidential visual reasoning, and permits users to trace error causes in the interpretable paths. We thus train CogCoM, a general 17B VLM with a memory-based compatible architecture endowed this reasoning mechanism. Experiments show that our model achieves the state-of-the-art performance across 8 benchmarks from 3 categories, and a limited number of training steps with the data swiftly gains a competitive performance. The code and data are publicly available at https://github.com/THUDM/CogCoM.

LongAlign: A Recipe for Long Context Alignment of Large Language Models

Extending large language models to effectively handle long contexts requires instruction fine-tuning on input sequences of similar length. To address this, we present LongAlign -- a recipe of the instruction data, training, and evaluation for long context alignment. First, we construct a long instruction-following dataset using Self-Instruct. To ensure the data diversity, it covers a broad range of tasks from various long context sources. Second, we adopt the packing and sorted batching strategies to speed up supervised fine-tuning on data with varied length distributions. Additionally, we develop a loss weighting method to balance the contribution to the loss across different sequences during packing training. Third, we introduce the LongBench-Chat benchmark for evaluating instruction-following capabilities on queries of 10k-100k in length. Experiments show that LongAlign outperforms existing recipes for LLMs in long context tasks by up to 30\%, while also maintaining their proficiency in handling short, generic tasks. The code, data, and long-aligned models are open-sourced at https://github.com/THUDM/LongAlign.

Text-Image Conditioned Diffusion for Consistent Text-to-3D Generation

By lifting the pre-trained 2D diffusion models into Neural Radiance Fields (NeRFs), text-to-3D generation methods have made great progress. Many state-of-the-art approaches usually apply score distillation sampling (SDS) to optimize the NeRF representations, which supervises the NeRF optimization with pre-trained text-conditioned 2D diffusion models such as Imagen. However, the supervision signal provided by such pre-trained diffusion models only depends on text prompts and does not constrain the multi-view consistency. To inject the cross-view consistency into diffusion priors, some recent works finetune the 2D diffusion model with multi-view data, but still lack fine-grained view coherence. To tackle this challenge, we incorporate multi-view image conditions into the supervision signal of NeRF optimization, which explicitly enforces fine-grained view consistency. With such stronger supervision, our proposed text-to-3D method effectively mitigates the generation of floaters (due to excessive densities) and completely empty spaces (due to insufficient densities). Our quantitative evaluations on the T$^3$Bench dataset demonstrate that our method achieves state-of-the-art performance over existing text-to-3D methods. We will make the code publicly available.

WaterBench: Towards Holistic Evaluation of Watermarks for Large Language Models

To mitigate the potential misuse of large language models (LLMs), recent research has developed watermarking algorithms, which restrict the generation process to leave an invisible trace for watermark detection. Due to the two-stage nature of the task, most studies evaluate the generation and detection separately, thereby presenting a challenge in unbiased, thorough, and applicable evaluations. In this paper, we introduce WaterBench, the first comprehensive benchmark for LLM watermarks, in which we design three crucial factors: (1) For \textbf{benchmarking procedure}, to ensure an apples-to-apples comparison, we first adjust each watermarking method's hyper-parameter to reach the same watermarking strength, then jointly evaluate their generation and detection performance. (2) For \textbf{task selection}, we diversify the input and output length to form a five-category taxonomy, covering $9$ tasks. (3) For \textbf{evaluation metric}, we adopt the GPT4-Judge for automatically evaluating the decline of instruction-following abilities after watermarking. We evaluate $4$ open-source watermarks on $2$ LLMs under $2$ watermarking strengths and observe the common struggles for current methods on maintaining the generation quality. The code and data are available at \url{https://github.com/THU-KEG/WaterBench}.

Know2BIO: A Comprehensive Dual-View Benchmark for Evolving Biomedical Knowledge Graphs

Knowledge graphs (KGs) have emerged as a powerful framework for representing and integrating complex biomedical information. However, assembling KGs from diverse sources remains a significant challenge in several aspects, including entity alignment, scalability, and the need for continuous updates to keep pace with scientific advancements. Moreover, the representative power of KGs is often limited by the scarcity of multi-modal data integration. To overcome these challenges, we propose Know2BIO, a general-purpose heterogeneous KG benchmark for the biomedical domain. Know2BIO integrates data from 30 diverse sources, capturing intricate relationships across 11 biomedical categories. It currently consists of ~219,000 nodes and ~6,200,000 edges. Know2BIO is capable of user-directed automated updating to reflect the latest knowledge in biomedical science. Furthermore, Know2BIO is accompanied by multi-modal data: node features including text descriptions, protein and compound sequences and structures, enabling the utilization of emerging natural language processing methods and multi-modal data integration strategies. We evaluate KG representation models on Know2BIO, demonstrating its effectiveness as a benchmark for KG representation learning in the biomedical field. Data and source code of Know2BIO are available at https://github.com/Yijia-Xiao/Know2BIO/.

T$^3$Bench: Benchmarking Current Progress in Text-to-3D Generation

Recent methods in text-to-3D leverage powerful pretrained diffusion models to optimize NeRF. Notably, these methods are able to produce high-quality 3D scenes without training on 3D data. Due to the open-ended nature of the task, most studies evaluate their results with subjective case studies and user experiments, thereby presenting a challenge in quantitatively addressing the question: How has current progress in Text-to-3D gone so far? In this paper, we introduce T$^3$Bench, the first comprehensive text-to-3D benchmark containing diverse text prompts of three increasing complexity levels that are specially designed for 3D generation. To assess both the subjective quality and the text alignment, we propose two automatic metrics based on multi-view images produced by the 3D contents. The quality metric combines multi-view text-image scores and regional convolution to detect quality and view inconsistency. The alignment metric uses multi-view captioning and Large Language Model (LLM) evaluation to measure text-3D consistency. Both metrics closely correlate with different dimensions of human judgments, providing a paradigm for efficiently evaluating text-to-3D models. The benchmarking results, shown in Fig. 1, reveal performance differences among six prevalent text-to-3D methods. Our analysis further highlights the common struggles for current methods on generating surroundings and multi-object scenes, as well as the bottleneck of leveraging 2D guidance for 3D generation. Our project page is available at: https://t3bench.com.