AWhat changed
Developed general-domain LM pretraining, target-domain adaptation and classifier fine-tuning using discriminative layerwise rates, slanted triangular schedules, gradual unfreezing and pooled recurrent states.
Scientific scope
Developed general-domain LM pretraining, target-domain adaptation and classifier fine-tuning using discriminative layerwise rates, slanted triangular schedules, gradual unfreezing and pooled recurrent states. Januaryv1 is FitLaM with five datasets; finalACL paper is ULMFiT with six, retained as one contribution rather than two methods. Inherits LM fine-tuning, AWD-LSTM, triangular rates and chain-thaw-style gradual adaptation. Universal names practical reuse criteria, not proof for every NLP task or100fold universal sample efficiency; pretraining and tuning costs remain.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Januaryv1 is FitLaM with five datasets; finalACL paper is ULMFiT with six, retained as one contribution rather than two methods. Inherits LM fine-tuning, AWD-LSTM, triangular rates and chain-thaw-style gradual adaptation. Universal names practical reuse criteria, not proof for every NLP task or100fold universal sample efficiency; pretraining and tuning costs remain.
Contributors
- Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.
- Fine-tuned Language Models for Text Classification
- Universal Language Model Fine-tuning for Text Classification
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction