AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Can Large Language Models Generalize Procedures Across Representations?

Fangru Lin, Valentin Hofmann, Xingchen Wan, Weixing Wang, Zifeng Ding, Anthony G Cohn, Janet B. Pierrehumbert

ICML 2026 · Large language models › Language, safety and interpretability · approach: LLM unlearning and benchmarks · 0 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
add Graph stage-1 GRPO training before NL trainingQwen2.5-1.5B-Instruct GRPO · Asynchronous planning NL testaccuracy: 0.698 -> 0.782Improved
reverse curriculum order to NL then Graph (ablation)Graph->NL curriculum · Asynchronous planning NL testaccuracy: 0.698 -> 0.431Hurt
replace stage 2 GRPO with NL distillation (ablation)Graph->NL curriculum · Asynchronous planning NL testaccuracy: 0.782 -> 0.462Hurt
apply the curriculum with standard SFT (ablation)Qwen2.5-1.5B-Instruct SFT · Asynchronous planning NL testaccuracy: 0.236 -> 0.249No clear effect
interleave Graph and NL data instead of sequential curriculum (ablation)Graph->NL curriculum · Asynchronous planning NL testaccuracy: 0.782 -> 0.382Hurt
use Code instead of Graph as stage 1 (ablation)symbolic-then-NL curriculum · Asynchronous planning NL testaccuracy: 0.782 -> 0.382Hurt
mix Graph and Code in stage 1 (ablation)Graph->NL curriculum · Asynchronous planning NL testaccuracy: 0.782 -> 0.533Hurt
add Graph stage-1 curriculum before NL trainingQwen2.5-1.5B-Instruct GRPO · NL-AAVE dialect testaccuracy: 0.507 -> 0.573Improved
add Code stage-1 curriculum before NL training on mathQwen2.5-1.5B-Instruct GRPO · SciBench physics test (trained on math)accuracy: 0.135 -> 0.230Improved
add Code stage-1 curriculum before NL training on physicsQwen2.5-1.5B-Instruct GRPO · SciBench physics test (in-domain)accuracy: 0.555 -> 0.550No clear effect
scale model size from 1.5B to 7B under NL-only trainingQwen2.5 GRPO NL training · Asynchronous planning NL testaccuracy: 0.698 -> 0.698No clear effect
train on Graph or Code only, test on NLSFT and GRPO post-training · Asynchronous planning NL testaccuracyNo clear effect

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