AI Paper Atlas

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

EarthSE: A Benchmark Evaluating Earth Scientific Exploration Capability for Large Language Models

Wanghan Xu, Xiangyu Zhao, Yuhao Zhou, Xiaoyu Yue, Ben Fei, Fenghua Ling, Wenlong Zhang, LEI BAI

ICLR 2026 · Large language models › Agents, code and math · approach: Mathematical and structural reasoning benchmarks

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
provide 3 initial CoT steps as hints at inference (ablation)DeepSeek-V3 · Earth-Silver FIBFIB accuracy (%): 12.80 -> 45.60Improved
provide 3 initial CoT steps as hints at inference (ablation)GPT-4o · Earth-Silver FIBFIB accuracy (%): 18.40 -> 55.60Improved
provide 1 initial CoT step (ablation)DeepSeek-V3 · Earth-Silver FIBFIB accuracy (%): 12.80 -> 21.60Improved
provide 2 initial CoT steps (ablation)GPT-4o · Earth-Silver FIBFIB accuracy (%): 18.40 -> 38.80Improved
move from Earth-Iron to harder Earth-Silvermean of 11 LLMs · Multiple choiceaccuracy (%): 89.53 -> 54.40Hurt
move from Earth-Iron to Earth-Silvermean of 11 LLMs · Free responsewin rate (%): 72.86 -> 41.15Hurt
scale Llama-3.1 from 8B to 70BLlama-3.1 · Earth-Iron MCaccuracy (%): 59.41 -> 91.56Improved
scale Llama-3.1 from 8B to 70BLlama-3.1 · Earth-GoldSES: 0.3301 -> 0.2453Hurt

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