Pith. sign in

REVIEW 2 cited by

Optimizing Deeper Transformers on Small Datasets

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2012.15355 v4 pith:GXGTFHA6 submitted 2020-12-30 cs.CL cs.LG

Optimizing Deeper Transformers on Small Datasets

classification cs.CL cs.LG
keywords datasetslayerssmalltransformerschallengingdeepinitializationparsing
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
abstract

It is a common belief that training deep transformers from scratch requires large datasets. Consequently, for small datasets, people usually use shallow and simple additional layers on top of pre-trained models during fine-tuning. This work shows that this does not always need to be the case: with proper initialization and optimization, the benefits of very deep transformers can carry over to challenging tasks with small datasets, including Text-to-SQL semantic parsing and logical reading comprehension. In particular, we successfully train $48$ layers of transformers, comprising $24$ fine-tuned layers from pre-trained RoBERTa and $24$ relation-aware layers trained from scratch. With fewer training steps and no task-specific pre-training, we obtain the state-of-the-art performance on the challenging cross-domain Text-to-SQL parsing benchmark Spider. We achieve this by deriving a novel Data-dependent Transformer Fixed-update initialization scheme (DT-Fixup), inspired by the prior T-Fixup work. Further error analysis shows that increasing depth can help improve generalization on small datasets for hard cases that require reasoning and structural understanding.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. H$_2$O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models

    cs.LG 2023-06 unverdicted novelty 6.0

    H2O evicts non-heavy-hitter tokens from the KV cache using a dynamic submodular policy, retaining recent and frequent-co-occurrence tokens to reduce memory while preserving accuracy.

  2. Rethinking the long-range dependency in Mamba/SSM and transformer models

    cs.LG 2025-09 reject novelty 3.0

    SSM/Mamba long-range dependency decays exponentially with the time gap by construction; a proposed interaction-based hidden state update can break this decay, but its proven stability covers only a restrictive special case.