AI Paper Atlas

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

Hypothesis-Driven Reasoning for Large Language Models

Aakash Kumar Agarwal, Moyuru Yamada

AAAI 2026 · Large language models › Agents, code and math · approach: Mathematical and structural reasoning 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 oracle hypotheses (semantic memory) instead of raw episodes onlyGPT-4o · Experience-Transfer Lv.2 (no CoT)accuracy (%): 35.3 -> 92.0Improved
add oracle semantic memory to raw episodesGemma3-12B · Experience-Transfer Lv.1 (no CoT)accuracy (%): 18.0 -> 56.7Improved
add oracle semantic memory alongside episodic memoryGPT-4o · Experience-Transfer Lv.3 (CoT)accuracy (%): 55.3 -> 66.0 (but semantic alone 68.7)Mixed
replace per-episode Reflexion with HDRHDR · Experience-Transfer Lv.2 (Gemma3-12B)accuracy (%): 26.0 -> 80.0Improved
replace Reflexion (individual) with HDRHDR · Experience-Transfer Lv.2 (GPT-4o)accuracy (%): 59.3 -> 84.7Improved
HDR vs oracle hypothesesHDR · Experience-Transfer Lv.2 (GPT-4o)accuracy (%): 94.7 -> 84.7Hurt
add factor extraction to direct hypothesis (ablation)HDR · Experience-Transfer Lv.2hypothesis F1: 0.0 -> 1.0Improved
add iterative refinement to simple two-stage (ablation)HDR · Experience-Transfer Lv.3hypothesis F1: 0.36 -> 1.0Improved
use Gemini-2.5 Flash instead of GPT-4o for consolidation (ablation)HDR · Experience-Transfer avgaccuracy (%): 72.2 -> 76.2Improved
use Gemma3-12B for consolidation (ablation)HDR · Experience-Transfer Lv.1accuracy (%): 84.0 -> 90.7 (Lv.1), 84.7 -> 43.3 (Lv.2)Mixed

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