Can In-context Learners Learn a Reasoning Concept from Demonstrations?
Identifikátory výsledku
Kód výsledku v IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216224%3A14330%2F23%3A00131346" target="_blank" >RIV/00216224:14330/23:00131346 - isvavai.cz</a>
Výsledek na webu
—
DOI - Digital Object Identifier
—
Alternativní jazyky
Jazyk výsledku
angličtina
Název v původním jazyce
Can In-context Learners Learn a Reasoning Concept from Demonstrations?
Popis výsledku v původním jazyce
Language models exhibit an emergent ability to learn a new task from a small number of input-output demonstrations. However, recent work shows that in-context learners largely rely on their pre-trained knowledge, such as the sentiment of the labels, instead of learning new associations from the input. We argue that the commonly-used few-shot evaluation using a random selection of in-context demonstrations can not disentangle models' reliance on such biases, as most of the randomly-selected demonstrations do not present relations informative for prediction beyond exposing the task's input-output distribution. Therefore, to evaluate models' in-context learning ability independent of models' memory, we introduce a Concept-sharing few-shot learning method choosing the demonstrations that share an underlying concept with the predicted sample. We extract a set of such concepts from available human explanations and measure how much models can benefit from presenting these concepts in few-shot demonstrations. We find that most of the recent in-context learners can not consistently benefit from the demonstrated concepts, irrespective of the model size. However, we note that T0 models are more sensitive to exhibited concepts, benefiting from concept-sharing demonstrations in 7 out of 8 evaluation scenarios.
Název v anglickém jazyce
Can In-context Learners Learn a Reasoning Concept from Demonstrations?
Popis výsledku anglicky
Language models exhibit an emergent ability to learn a new task from a small number of input-output demonstrations. However, recent work shows that in-context learners largely rely on their pre-trained knowledge, such as the sentiment of the labels, instead of learning new associations from the input. We argue that the commonly-used few-shot evaluation using a random selection of in-context demonstrations can not disentangle models' reliance on such biases, as most of the randomly-selected demonstrations do not present relations informative for prediction beyond exposing the task's input-output distribution. Therefore, to evaluate models' in-context learning ability independent of models' memory, we introduce a Concept-sharing few-shot learning method choosing the demonstrations that share an underlying concept with the predicted sample. We extract a set of such concepts from available human explanations and measure how much models can benefit from presenting these concepts in few-shot demonstrations. We find that most of the recent in-context learners can not consistently benefit from the demonstrated concepts, irrespective of the model size. However, we note that T0 models are more sensitive to exhibited concepts, benefiting from concept-sharing demonstrations in 7 out of 8 evaluation scenarios.
Klasifikace
Druh
D - Stať ve sborníku
CEP obor
—
OECD FORD obor
10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)
Návaznosti výsledku
Projekt
—
Návaznosti
S - Specificky vyzkum na vysokych skolach
Ostatní
Rok uplatnění
2023
Kód důvěrnosti údajů
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů
Údaje specifické pro druh výsledku
Název statě ve sborníku
Proceedings of the 1st Workshop on Natural Language Reasoning and Structured Explanations (NLRSE)
ISBN
9781959429944
ISSN
0736-587X
e-ISSN
—
Počet stran výsledku
9
Strana od-do
107-115
Název nakladatele
The Association for Computational Linguistics
Místo vydání
Toronto, Canada
Místo konání akce
Toronto, Canada
Datum konání akce
1. 1. 2023
Typ akce podle státní příslušnosti
WRD - Celosvětová akce
Kód UT WoS článku
—