IC-425Pretrained GPT-2 Large fails at structural in-context learning on unseen tokens in a syllogism task

Suraj Anand, Michael A. Lepori, Jack Merullo, Ellie Pavlick

SourceDual Process Learning: Controlling Use of In-Context vs. In-Weights Strategies with Weight Forgetting

The paper evaluates pretrained GPT-2 Large on a syllogism task where the subject term is replaced by a randomly initialized embedding, requiring the model to rely purely on contextual structure rather than token identity. The model shows persistently low accuracy on this unseen-token condition, described as 'even worse than on the pretrained pythia 1.4b.' This demonstrates that a widely used released language model lacks the ability to perform structural ICL on novel tokens in a logical reasoning setting.

Evidence
observational
Caveat
The evaluation is on a single simple syllogism task with a-z tokens; the paper provides no specific accuracy number for the pretrained model, only a qualitative comparison to Pythia-1.4B.
Model
GPT-2 Large
Concepts
Failure mode
Datasets
WikiText [eval]
Related work
Lampinen et al. 2024 (content effects on reasoning) [context], Kim et al. 2024 (mechanistic interpretation of syllogistic reasoning) [context]
Related findings
IC-424, IC-426
Extraction
automatic-extraction