Thailand

Recent progress in large language models (LLMs) has marked a notable milestone in the field of artificial intelligence. The conventional evaluation of LLMs primarily relies on existing tasks and benchmarks, raising concerns about test set contamination and the genuine comprehension abilities of LLMs. To address these concerns, we propose to evaluate LLMs by designing new tasks, automatically generating evaluation datasets for the tasks, and conducting detailed error analyses to scrutinize LLMs&#39; adaptability to new tasks, their sensitivity to prompt variations, and their error tendencies. We investigate the capacity of LLMs to adapt to new but simple tasks, especially when they diverge from the models&#39; pre-existing knowledge. Our methodology emphasizes the creation of straightforward tasks, facilitating a precise error analysis to uncover the underlying causes of LLM failures. This strategic approach also aims to uncover effective strategies for enhancing LLM performance based on the detailed error analysis of system output.

ACL 2024

Challenging Large Language Models with New Tasks: A Study on their Adaptability and Robustness

llm adaptability and robustness

llm evaluation with new tasks

error analysis

### Welcome!
The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. Our Virtual Poster Sessions will take place online Thursday, August 22, 2024.

You are required to register for this event. **Please register [here](https://2024.aclweb.org/registration). **

If you have already registered, please check your inbox for an email from Underline granting you access to ACL 2024 content.

Please register!

The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. More information will be announced soon.

Recent progress in large language models (LLMs) has marked a notable milestone in the field of artificial intelligence. The conventional evaluation of LLMs primarily relies on existing tasks and benchmarks, raising concerns about test set contamination and the genuine comprehension abilities of LLMs. To address these concerns, we propose to evaluate LLMs by designing new tasks, automatically generating evaluation datasets for the tasks, and conducting detailed error analyses to scrutinize LLMs' adaptability to new tasks, their sensitivity to prompt variations, and their error tendencies. We investigate the capacity of LLMs to adapt to new but simple tasks, especially when they diverge from the models' pre-existing knowledge. Our methodology emphasizes the creation of straightforward tasks, facilitating a precise error analysis to uncover the underlying causes of LLM failures. This strategic approach also aims to uncover effective strategies for enhancing LLM performance based on the detailed error analysis of system output.

poster

Advances in model editing through neuron pruning hold promise for removing undesirable concepts from large language models. However, it remains unclear whether models have the capacity to reacquire pruned concepts after editing. To investigate this, we evaluate concept relearning in models by tracking concept saliency and similarity in pruned neurons during retraining for named entity recognition tasks. Our findings reveal that models can quickly regain performance post-pruning by relocating advanced concepts to earlier layers and reallocating pruned concepts to primed neurons with similar semantics. This suggests that models exhibit polysemantic capacities and can blend old and new concepts in individual neurons. While neuron pruning provides interpretability into model concepts, our results highlight the challenges of permanent concept removal for improved model *safety*. Monitoring concept reemergence and developing techniques to mitigate relearning of unsafe concepts will be important directions for more robust model editing. Overall, our work strongly demonstrates the resilience and fluidity of concept representations in LLMs post concept removal.

Large Language Models Relearn Removed Concepts

Large language models show compelling performance on reasoning tasks but they tend to perform much worse in languages other than English. This is unsurprising given that their training data largely consists of English text and instructions. A typical solution is to translate instruction data into all languages of interest, and then train on the resulting multilingual data, which is called translate-training. This approach not only incurs high cost, but also results in poorly translated data due to the non-standard formatting of mathematical chain-of-thought. In this paper, we explore the benefits of question alignment, where we train the model to translate reasoning questions into English by finetuning on X-English parallel question data. In this way we perform targeted, in-domain language alignment which makes best use of English instruction data to unlock the LLMs' multilingual reasoning abilities. Experimental results on LLaMA2-13B show that question alignment leads to consistent improvements over the translate-training approach: an average improvement of 11.3% and 16.1% accuracy across ten languages on the MGSM and MSVAMP multilingual reasoning benchmarks.

Question Translation Training for Better Multilingual Reasoning

For a LLM to be trustworthy, its confidence level should be well-calibrated with its actual performance. While it is now common sense that LLM performances are greatly impacted by prompts, the confidence calibration in prompting LLMs has yet to be thoroughly explored.
In this paper, we explore how different prompting strategies influence LLM confidence calibration and how it could be improved. We conduct extensive experiments on six prompting methods in the question-answering context and we observe that, while these methods help improve the expected LLM calibration, they also trigger LLMs to be over-confident when responding to some instances.
Inspired by human cognition, we propose Fact-and-Reflection (FaR) prompting, which improves the LLM calibration in two steps. First, FaR elicits the known "facts" that are relevant to the input prompt from the LLM. And then it asks the model to "reflect" over them to generate the final answer.
Experiments show that FaR prompting achieves significantly better calibration; it lowers the Expected Calibration Error by 23.5% on our multi-purpose QA tasks. 
Notably, FaR prompting even elicits the capability of verbally expressing concerns in less confident scenarios, which helps trigger retrieval augmentation for solving these harder instances.

Fact-and-Reflection (FaR) Improves Confidence Calibration of Large Language Models

Recently, there is a surge in interest surrounding video large language models (Video LLMs). However, existing benchmarks fail to provide a comprehensive feedback on the temporal perception ability of Video LLMs. On the one hand, most of them are unable to distinguish between different temporal aspects (e.g., speed, direction) and thus cannot reflect the nuanced performance on these specific aspects. On the other hand, they are limited in the diversity of task formats (e.g., only multi-choice QA), which hinders the understanding of how temporal perception performance may vary across different types of tasks. Motivated by these two problems, we propose the \textbf{TempCompass} benchmark, which introduces a diversity of temporal aspects and task formats. To collect high-quality test data, we devise two novel strategies: (1) In video collection, we construct conflicting videos that share the same static content but differ in a specific temporal aspect, which prevents Video LLMs from leveraging single-frame bias or language priors. (2) To collect the task instructions, we propose a paradigm where humans first annotate meta-information for a video and then an LLM generates the instruction. We also design an LLM-based approach to automatically and accurately evaluate the responses from Video LLMs. Based on TempCompass, we comprehensively evaluate 9 state-of-the-art (SOTA) Video LLMs and 3 Image LLMs, and reveal the discerning fact that these models exhibit notably poor temporal perception ability.

TempCompass: Do Video LLMs Really Understand Videos?

Existing large language models struggle to support numerous low-resource languages, particularly the extremely low-resource ones, for which there is minimal training data available for effective parameter updating. We thus investigate whether LLMs can learn a new language on the fly solely through prompting. To study this question, we collect a research suite for Zhuang, a language supported by no LLMs currently. We introduce DiPMT++, a framework for adapting LLMs to unseen languages by in-context learning. Using a dictionary and 5K parallel sentences only, DiPMT++ significantly enhances the performance of GPT-4 from 0 to 16 BLEU for Chinese-to-Zhuang translation and achieves 32 BLEU for Zhuang-to-Chinese translation. We also validate the effectiveness of our framework on Kalamang, another unseen language. Furthermore, we demonstrate the practical utility of DiPMT++ in aiding humans in translating completely unseen languages, which could contribute to the preservation of linguistic diversity.

Teaching Large Language Models an Unseen Language on the Fly

We present DiffChat, a novel method to align Large Language Models (LLMs) to ``chat'' with prompt-as-input 
Text-to-Image Synthesis (TIS)
models (e.g., Stable Diffusion) for interactive image creation. 
Given a raw prompt/image and a user-specified instruction, DiffChat can effectively make appropriate modifications and generate the target prompt, which can be leveraged to create the target image of high quality. To achieve this, we first collect an instruction-following prompt engineering dataset named InstructPE for the supervised training of DiffChat.
Next, we propose a reinforcement learning framework with the feedback of three core criteria for image creation, i.e., aesthetics, user preference and content integrity. 
It involves an 
action-space dynamic modification technique to obtain more relevant positive samples and harder negative samples during the off-policy sampling. 
Content integrity is also introduced into the value estimation function for further improvement of produced images. 
Our method can exhibit superior performance than baseline models and strong competitors based on both automatic and human evaluations, which fully demonstrates its effectiveness.

DiffChat: Learning to Chat with Text-to-Image Synthesis Models for Interactive Image Creation

Multi-modal Event Reasoning (MMER) endeavors to endow machines with the ability to comprehend intricate event relations across diverse data modalities. MMER is fundamental and underlies a wide broad of applications. Despite extensive instruction fine-tuning, current multi-modal large language models still fall short in such ability. The disparity stems from that existing models are insufficient to capture underlying principles governing event evolution in various scenarios. In this paper, we introduce Multi-Modal Event Evolution Learning (MEEL) to enable the model to grasp the event evolution mechanism yielding advanced MMER ability. Specifically, we commence with the design of event diversification to gather seed events from a rich spectrum of scenarios. Subsequently, we employ ChatGPT to generate evolving graphs for these seed events. We propose an instruction encapsulation process that formulates the evolving graphs into instruction-tuning data, aligning the comprehension of event reasoning to humans. Finally, we observe that models trained in this way are still struggling to fully comprehend event evolution. In such a case, we propose the guiding discrimination strategy, in which models are trained to discriminate the improper evolution direction. We collect and curate a benchmark M-EV2 for MMER. Extensive experiments on M-EV2 validate the effectiveness of our approach, showcasing competitive performance in open-source multi-modal LLMs.

MEEL: Multi-Modal Event Evolution Learning

Large language models with instruction-following abilities have revolutionized the field of artificial intelligence. These models show exceptional generalizability to tackle various real-world tasks through their natural language interfaces. However, their performance heavily relies on high-quality exemplar data, which is often difficult to obtain. This challenge is further exacerbated when it comes to multimodal instruction following. We introduce TextBind, an almost annotation-free framework for empowering LLMs with multi-turn interleaved multimodal instruction-following capabilities. Our approach requires only image-caption pairs and generates multi-turn multimodal instruction-response conversations from a language model. To accommodate interleaved image-text inputs and outputs, we devise MIM, a language model-centric architecture that seamlessly integrates image encoder and decoder models. Extensive quantitative and qualitative experiments demonstrate that MIM trained on TextBind achieves remarkable generation capability in multimodal conversations compared to recent baselines.

TextBind: Multi-turn Interleaved Multimodal Instruction-following in the Wild

We present Q-ViD, a simple approach for video question answering (video QA), that unlike prior methods, which are based on complex architectures, computationally expensive pipelines or use closed models like GPTs, Q-ViD relies on a single instruction-aware open vision-language model (InstructBLIP) to tackle videoQA using frame descriptions. Specifically, we create captioning instruction prompts that rely on the target questions about the videos and leverage InstructBLIP to obtain video frame captions that are useful to the task at hand. Subsequently, we form descriptions of the whole video using the question-dependent frame captions, and feed that information, along with a question-answering prompt, to a large language model (LLM). The LLM is our reasoning module, and performs the final step of multiple-choice QA. Our simple Q-ViD framework achieves competitive or even higher performances than current state of the art models on a diverse range of videoQA benchmarks, including NExT-QA, STAR, How2QA, TVQA and IntentQA.

Question-Instructed Visual Descriptions for Zero-Shot Video Answering

Through pretraining on a corpus with various sources, Large Language Models (LLMs) have gained impressive performance. However, the impact of each component of the pretraining corpus remains opaque. As a result, the organization of the pretraining corpus is still empirical and may deviate from the optimal. To address this issue, we systematically analyze the impact of 48 datasets from 5 major categories of pretraining data of LLMs and measure their impacts on LLMs using benchmarks about nine major categories of model capabilities. Our analyses provide empirical results about the contribution of multiple corpora on the performances of LLMs, along with their joint impact patterns, including complementary, orthogonal, and correlational relationships. We also identify a set of ``high-impact data'' such as Books that is significantly related to a set of model capabilities. These findings provide insights into the organization of data to support more efficient pretraining of LLMs.

Downloads

Next from ACL 2024

Large Language Models Relearn Removed Concepts

Similar lecture

Pareto Optimal Learning for Estimating Large Language Model Errors