AI Paper Atlas

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

Can Language Models Discover Scaling Laws?

Haowei Lin, Haotian Ye, Wenzheng Feng, Quzhe Huang, Yujun Li, Hubert Lim, Zhengrui Li, Xiangyu Wang, Jianzhu Ma, Yitao Liang, James Zou

ICLR 2026 · Large language models › Agents, code and math · approach: Agent instruction-following benchmarks · 12 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 SLDAgent evolutionary co-optimization of law and optimizer instead of native agent scaffoldGPT-5 / Codex · SLDBench avgR^2: 0.550 -> 0.748Improved
use SLDAgent instead of Gemini-CLIGemini-2.5-Flash · SLDBench avgR^2: 0.077 -> 0.506Improved
use SLDAgent instead of Claude CodeClaude-Haiku-4.5 · SLDBench avgR^2: 0.419 -> 0.519Improved
use SLDAgent instead of Gemini-CLIGemini-3-Pro-Preview · SLDBench avgR^2: 0.382 -> 0.636Improved
use SLDAgent instead of Codexo4-mini · SLDBench avgR^2: 0.349 -> 0.657Improved
SLDAgent vs human-derived lawSLDAgent GPT-5 · lr&bszR^2: -0.076 -> 0.604Improved
SLDAgent vs human-derived lawSLDAgent GPT-5 · u_shapeR^2: -1.000 -> -0.305Improved
SLDAgent vs Goose on u_shape and moeSLDAgent GPT-5 · moeR^2: 0.813 -> 0.773Hurt
SLDAgent with weak models on lr&bszSLDAgent (G-2.5-Flash, C-Haiku-4.5) · lr&bszR^2: -0.873 -> -0.871; -0.511 -> -0.657Mixed
SLDAgent-discovered moe law vs human lawSLDAgent · moeR^2: 0.732 -> 0.891Improved
use SLDAgent law for fine-tuning model selection vs SubTuningLLM selection · WikitextPearCorr: 90.3 -> 83.6Hurt
use SLDAgent law for LLM selection vs SubTuningLLM selection · FLAN, GigawordPearCorr: 79.4 -> 84.6; 86.3 -> 94.8Improved

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