AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

RECKONING: Reasoning through Dynamic Knowledge Encoding

Zeming Chen, Gail Weiss, Eric Mitchell, Asli Celikyilmaz, Antoine Bosselut

NeurIPS 2023 · Large language models › Retrieval and knowledge · approach: Knowledge graph reasoning with LLMs · 5 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
Encode contextual knowledge in model parameters using gradient updates, then answer from the updated modelRECKONING · ProofWriter and CLUTRR-SGlabel accuracy: 99.4 -> 99.5; 99.2 -> 99.7; 99.6 -> 99.8; 98.1 -> 98.3; 96.9 -> 97.6; 90.3 -> 94.8Improved
Add a knowledge-recovery CLM objective to question-answering training (ablation)RECKONING · ProofWriter and CLUTRR-SGlabel accuracy: 99.4 -> 99.5; 98.3 -> 99.5; 98.3 -> 99.7; 99.1 -> 99.8; 96.0 -> 98.3; 90.2 -> 94.8Improved
Add a knowledge-recovery objective to RECKONING training (ablation)RECKONING · ProofWriter with all distractorslabel accuracy: ~18.2%Improved
Replace dynamic per-step, per-layer learning rates with one static shared learning rate (ablation)RECKONING · CLUTRR-SGlabel accuracy: average drop of 34.2%; 45.5% drop for 4-hop; 39.5% drop for 6-hopHurt
Increase inner-loop knowledge-encoding steps from one to six (ablation)RECKONING · CLUTRR-SGlabel accuracy: 5.9% for 2-hop; 42.3% for 4-hop; 34.7% for 6-hopImproved
Add all distractor facts to the knowledge context (ablation)FT-ICR · ProofWriterlabel accuracy: 28.6%Hurt
Add all distractor facts to the knowledge context (ablation)RECKONING · ProofWriterlabel accuracy: 17.1%Hurt
Use RECKONING rather than FT-ICR when all distractors are presentGPT-2-small · ProofWriter with distractorsaverage label accuracy: 70.9% -> 82.5%Improved
Use RECKONING rather than in-context reasoningGPT-2-XL with LoRA · ProofWriter 5-hop with all distractorsaccuracy: 65% -> 70.2%Improved
Encode shared knowledge once and reuse updated parameters for multiple questionsRECKONING 1-step · Synthetic random-letter task, 18 questionswall-clock time (s): 2.0436 -> 0.6228Improved
Encode shared knowledge once and reuse updated parameters for multiple questionsRECKONING 4-step · Synthetic random-letter task, 18 questionswall-clock time (s): 2.0436 -> 1.4839Improved
Use RECKONING rather than GPT-2 in-context reasoningFOLIO · FOLIOaccuracy: 53.3 -> 54.9Improved

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