AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

TART: A plug-and-play Transformer module for task-agnostic reasoning

Kush Bhatia, Avanika Narayan, Christopher De Sa, Christopher Re

NeurIPS 2023 · Large language models › Instruction and fine-tuning · approach: Calibration and knowledge in LLM tuning · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use TART reasoning module on top of base LLM embeddingsLLM in-context learning · 14 NLP tasksaccuracy: +18.4 pointsImproved
TART vs adapter headsTART · 14 NLP tasksaccuracy: +3.4 pointsImproved
TART vs full fine-tuningTART · 14 NLP tasksaccuracy: within 3.1 pointsHurt
vanilla to leave-one-out (LOO) embeddings (ablation)TART · NLP tasksaccuracyImproved
scale in-context examples from 18 to 64TART · 14 NLP tasksaccuracy: 4.8% vs 9.0% for full fine-tuningImproved
scale base model from 100M to 6BTART · GPT-Neo familyaccuracy: +9.8%Improved
scale base model from 100M to 6BTART vs fine-tuning · GPT-Neo familyaccuracy gap: 7.5% -> 2.2%Improved
scale in-context examples from 18 to 256TART · long-text NLP tasksaccuracy: +6.8%Improved
TART vs in-context learning beyond context limitTART · long-text NLP tasksaccuracy: 19.1%Improved
TART with GPT-Neo 125M vs Bloom 176BTART · RAFTaccuracy: 0.595 -> 0.634Improved
TART with GPT-Neo 125M vs GPT-3 175BTART · RAFTaccuracy: 0.673 -> 0.634Hurt
task-specific fine-tuning on AGNewsLLM · other tasksaccuracy: 25.77%Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy