Identifying Semantic Induction Heads to Understand In-Context Learning
- URL: http://arxiv.org/abs/2402.13055v2
- Date: Thu, 25 Jul 2024 08:07:39 GMT
- Title: Identifying Semantic Induction Heads to Understand In-Context Learning
- Authors: Jie Ren, Qipeng Guo, Hang Yan, Dongrui Liu, Quanshi Zhang, Xipeng Qiu, Dahua Lin,
- Abstract summary: We investigate whether attention heads encode two types of relationships between tokens present in natural languages.
We find that certain attention heads exhibit a pattern where, when attending to head tokens, they recall tail tokens and increase the output logits of those tail tokens.
- Score: 103.00463655766066
- License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
- Abstract: Although large language models (LLMs) have demonstrated remarkable performance, the lack of transparency in their inference logic raises concerns about their trustworthiness. To gain a better understanding of LLMs, we conduct a detailed analysis of the operations of attention heads and aim to better understand the in-context learning of LLMs. Specifically, we investigate whether attention heads encode two types of relationships between tokens present in natural languages: the syntactic dependency parsed from sentences and the relation within knowledge graphs. We find that certain attention heads exhibit a pattern where, when attending to head tokens, they recall tail tokens and increase the output logits of those tail tokens. More crucially, the formulation of such semantic induction heads has a close correlation with the emergence of the in-context learning ability of language models. The study of semantic attention heads advances our understanding of the intricate operations of attention heads in transformers, and further provides new insights into the in-context learning of LLMs.
Related papers
- Interpreting Context Look-ups in Transformers: Investigating Attention-MLP Interactions [19.33740818235595]
This study investigates how attention heads and next-token neurons interact in large language models (LLMs) to predict new words.
Our findings reveal that some attention heads recognize specific contexts and activate a token-predicting neuron accordingly.
arXiv Detail & Related papers (2024-02-23T02:15:47Z) - From Heuristic to Analytic: Cognitively Motivated Strategies for
Coherent Physical Commonsense Reasoning [66.98861219674039]
Heuristic-Analytic Reasoning (HAR) strategies drastically improve the coherence of rationalizations for model decisions.
Our findings suggest that human-like reasoning strategies can effectively improve the coherence and reliability of PLM reasoning.
arXiv Detail & Related papers (2023-10-24T19:46:04Z) - Improving Open Information Extraction with Large Language Models: A
Study on Demonstration Uncertainty [52.72790059506241]
Open Information Extraction (OIE) task aims at extracting structured facts from unstructured text.
Despite the potential of large language models (LLMs) like ChatGPT as a general task solver, they lag behind state-of-the-art (supervised) methods in OIE tasks.
arXiv Detail & Related papers (2023-09-07T01:35:24Z) - IERL: Interpretable Ensemble Representation Learning -- Combining
CrowdSourced Knowledge and Distributed Semantic Representations [11.008412414253662]
Large Language Models (LLMs) encode meanings of words in the form of distributed semantics.
Recent studies have shown that LLMs tend to generate unintended, inconsistent, or wrong texts as outputs.
We propose a novel ensemble learning method, Interpretable Ensemble Representation Learning (IERL), that systematically combines LLM and crowdsourced knowledge representations.
arXiv Detail & Related papers (2023-06-24T05:02:34Z) - In-Context Analogical Reasoning with Pre-Trained Language Models [10.344428417489237]
We explore the use of intuitive language-based abstractions to support analogy in AI systems.
Specifically, we apply large pre-trained language models (PLMs) to visual Raven's Progressive Matrices ( RPM)
We find that PLMs exhibit a striking capacity for zero-shot relational reasoning, exceeding human performance and nearing supervised vision-based methods.
arXiv Detail & Related papers (2023-05-28T04:22:26Z) - Simple Linguistic Inferences of Large Language Models (LLMs): Blind Spots and Blinds [59.71218039095155]
We evaluate language understanding capacities on simple inference tasks that most humans find trivial.
We target (i) grammatically-specified entailments, (ii) premises with evidential adverbs of uncertainty, and (iii) monotonicity entailments.
The models exhibit moderate to low performance on these evaluation sets.
arXiv Detail & Related papers (2023-05-24T06:41:09Z) - ChatABL: Abductive Learning via Natural Language Interaction with
ChatGPT [72.83383437501577]
Large language models (LLMs) have recently demonstrated significant potential in mathematical abilities.
LLMs currently have difficulty in bridging perception, language understanding and reasoning capabilities.
This paper presents a novel method for integrating LLMs into the abductive learning framework.
arXiv Detail & Related papers (2023-04-21T16:23:47Z) - ERICA: Improving Entity and Relation Understanding for Pre-trained
Language Models via Contrastive Learning [97.10875695679499]
We propose a novel contrastive learning framework named ERICA in pre-training phase to obtain a deeper understanding of the entities and their relations in text.
Experimental results demonstrate that our proposed ERICA framework achieves consistent improvements on several document-level language understanding tasks.
arXiv Detail & Related papers (2020-12-30T03:35:22Z)
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.