AI Paper Atlas

Research map / Large language models / Agents, code and math

How to transfer algorithmic reasoning knowledge to learn new algorithms?

Louis-Pascal A. C. Xhonneux, Andreea Deac, Petar Veličković, Jian Tang

NeurIPS 2021 · Large language models › Agents, code and math · approach: LLM-guided program synthesis · 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
use NE++ (more expressive encoder) instead of NE (ablation)NeuralExecutor · Dijkstra teacher forcing, 100 nodesPredecessor error: 0.064 -> 0.522Hurt
use NE++ instead of NE (ablation)NeuralExecutor · Most reliable teacher forcingNext node error: 0.238 -> 0.174Improved
use NE++ instead of NE (ablation)NeuralExecutor · Most reliable teacher forcing, 100 nodesPredecessor error: 0.167 -> 0.171No clear effect
remove intermediate steps (no algorithm) vs teacher forcing (ablation)NE · Dijkstra, 100 nodesKey MSE: 4.79 -> 3.34e16Hurt
fine-tune pretrained Prim modelNE · Dijkstra no algorithm, 100 nodesPredecessor error: 0.194 -> 0.388Hurt
freeze pretrained Prim processorNE · Dijkstra no algorithm, 100 nodesPredecessor error: 0.194 -> 0.607Hurt
2-Processor transfer from PrimNE · Dijkstra no algorithm, 100 nodesPredecessor error: 0.194 -> 0.305Hurt
2-Processor transfer from PrimNE · Dijkstra no algorithm, 100 nodesKey MSE: 3.34e16 -> 1.04e4Improved
multi-task training with PrimNE · Dijkstra, 100 nodesKey MSE: 3.34e16 -> 126Improved
multi-task training with PrimNE++ · Dijkstra, 100 nodesPredecessor error: 0.388 -> 0.282Improved
multi-task training with WidestNE++ · Most reliable, 50 nodesPredecessor error: 0.566 -> 0.185Improved
multi-task training with WidestNE · Most reliable, 20 nodesPredecessor error: 0.187 -> 0.452Hurt

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