Quantifying Generalization Complexity for Large Language Models

• URL: http://arxiv.org/abs/2410.01769v2
• Date: Thu, 3 Oct 2024 15:30:12 GMT
• Title: Quantifying Generalization Complexity for Large Language Models
• Authors: Zhenting Qi, Hongyin Luo, Xuliang Huang, Zhuokai Zhao, Yibo Jiang, Xiangjun Fan, Himabindu Lakkaraju, James Glass,
• Abstract summary: We introduce Scylla, a dynamic evaluation framework that quantitatively measures the generalization abilities of large language models. Scylla disentangles generalization from memorization via assessing model performance on both in-distribution (ID) and out-of-distribution (OOD) data. We benchmark 28LLMs including both open-sourced models such as LLaMA and Qwen families, and close-sourced models like Claude and GPT.
• Score: 31.721781613271066
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: While large language models (LLMs) have shown exceptional capabilities in understanding complex queries and performing sophisticated tasks, their generalization abilities are often deeply entangled with memorization, necessitating more precise evaluation. To address this challenge, we introduce Scylla, a dynamic evaluation framework that quantitatively measures the generalization abilities of LLMs. Scylla disentangles generalization from memorization via assessing model performance on both in-distribution (ID) and out-of-distribution (OOD) data through 20 tasks across 5 levels of complexity. Through extensive experiments, we uncover a non-monotonic relationship between task complexity and the performance gap between ID and OOD data, which we term the generalization valley. Specifically, this phenomenon reveals a critical threshold - referred to as critical complexity - where reliance on non-generalizable behavior peaks, indicating the upper bound of LLMs' generalization capabilities. As model size increases, the critical complexity shifts toward higher levels of task complexity, suggesting that larger models can handle more complex reasoning tasks before over-relying on memorization. Leveraging Scylla and the concept of critical complexity, we benchmark 28LLMs including both open-sourced models such as LLaMA and Qwen families, and close-sourced models like Claude and GPT, providing a more robust evaluation and establishing a clearer understanding of LLMs' generalization capabilities.

Related papers

• ZebraLogic: On the Scaling Limits of LLMs for Logical Reasoning [92.76959707441954]
  We introduce ZebraLogic, a comprehensive evaluation framework for assessing LLM reasoning performance. ZebraLogic enables the generation of puzzles with controllable and quantifiable complexity. Our results reveal a significant decline in accuracy as problem complexity grows.
  arXiv  Detail & Related papers  (2025-02-03T06:44:49Z)
• CLR-Fact: Evaluating the Complex Logical Reasoning Capability of Large Language Models over Factual Knowledge [44.59258397967782]
  Large language models (LLMs) have demonstrated impressive capabilities across various natural language processing tasks. We present a systematic evaluation of state-of-the-art LLMs' complex logical reasoning abilities. We find that LLMs excel at reasoning over general world knowledge but face significant challenges with specialized domain-specific knowledge.
  arXiv  Detail & Related papers  (2024-07-30T05:40:32Z)
• Do Large Language Models Have Compositional Ability? An Investigation into Limitations and Scalability [12.349247962800813]
  Large language models (LLMs) have emerged as powerful tools for many AI problems. They exhibit remarkable in-context learning (ICL) capabilities. How they approach composite tasks remains an open and largely underexplored question.
  arXiv  Detail & Related papers  (2024-07-22T15:22:34Z)
• Benchmarking Complex Instruction-Following with Multiple Constraints Composition [72.82640456309821]
  How to evaluate the ability of complex instruction-following of large language models (LLMs) has become a critical research problem. Existing benchmarks mainly focus on modeling different types of constraints in human instructions while neglecting the composition of different constraints. We propose ComplexBench, a benchmark for comprehensively evaluating the ability of LLMs to follow complex instructions composed of multiple constraints.
  arXiv  Detail & Related papers  (2024-07-04T14:50:45Z)
• A Notion of Complexity for Theory of Mind via Discrete World Models [2.487142846438629]
  Theory of Mind (ToM) can be used to assess the capabilities of Large Language Models (LLMs) in complex scenarios where social reasoning is required. This work proposes a framework inspired by cognitive load theory to measure the complexity of ToM tasks.
  arXiv  Detail & Related papers  (2024-06-16T16:46:55Z)
• Adaptive-RAG: Learning to Adapt Retrieval-Augmented Large Language Models through Question Complexity [59.57065228857247]
  Retrieval-augmented Large Language Models (LLMs) have emerged as a promising approach to enhancing response accuracy in several tasks, such as Question-Answering (QA) We propose a novel adaptive QA framework, that can dynamically select the most suitable strategy for (retrieval-augmented) LLMs based on the query complexity. We validate our model on a set of open-domain QA datasets, covering multiple query complexities, and show that ours enhances the overall efficiency and accuracy of QA systems.
  arXiv  Detail & Related papers  (2024-03-21T13:52:30Z)
• Unveiling the Generalization Power of Fine-Tuned Large Language Models [81.70754292058258]
  We investigate whether fine-tuning affects the intrinsic generalization ability intrinsic to Large Language Models (LLMs) Our main findings reveal that models fine-tuned on generation and classification tasks exhibit dissimilar behaviors in generalizing to different domains and tasks. We observe that integrating the in-context learning strategy during fine-tuning on generation tasks can enhance the model's generalization ability.
  arXiv  Detail & Related papers  (2024-03-14T08:18:59Z)
• Parrot Mind: Towards Explaining the Complex Task Reasoning of Pretrained Large Language Models with Template-Content Structure [66.33623392497599]
  We show that a structure called template-content structure (T-C structure) can reduce the possible space from exponential level to linear level. We demonstrate that models can achieve task composition, further reducing the space needed to learn from linear to logarithmic.
  arXiv  Detail & Related papers  (2023-10-09T06:57:45Z)
• Can Large Language Models Understand Real-World Complex Instructions? [54.86632921036983]
  Large language models (LLMs) can understand human instructions, but struggle with complex instructions. Existing benchmarks are insufficient to assess LLMs' ability to understand complex instructions. We propose CELLO, a benchmark for evaluating LLMs' ability to follow complex instructions systematically.
  arXiv  Detail & Related papers  (2023-09-17T04:18:39Z)
• Model-agnostic Measure of Generalization Difficulty [7.183430740278161]
  We propose the first model-agnostic measure of the inherent generalization difficulty of tasks. Our measure quantifies the total information required to generalize well on a task minus the information provided by the data. It scales exponentially with the intrinsic dimensionality of the space over which the model must generalize but only intuitively in resolution per dimension.
  arXiv  Detail & Related papers  (2023-05-01T18:48:55Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.