AI Paper Atlas

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

Text2Arch: A Dataset for Generating Scientific Architecture Diagrams from Natural Language Descriptions

Shivank Garg, Sankalp Mittal, Manish Gupta

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Controlled text generation · 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
fine-tune DeepSeek-7B on Text2Arch instead of few-shot promptingDeepSeek-7B · Text2Arch test setNode F1: 43.5 -> 65.7Improved
fine-tune Llama-3-8B on Text2Arch instead of few-shot promptingLlama-3-8B · Text2Arch test setNode F1: 59.7 -> 31.9Hurt
fine-tune Qwen2-7B on Text2Arch instead of few-shot promptingQwen2-7B · Text2Arch test setEdge F1: 24.3 -> 13.7Hurt
fine-tune DeepSeek-7B instead of few-shot promptingDeepSeek-7B · Text2Arch test setEdge F1: 16.6 -> 38.0Improved
fine-tune DeepSeek-7B instead of few-shot promptingDeepSeek-7B · Text2Arch manual annotation set (99 images)Edge F1: 20.4 -> 49.1Improved
use fine-tuned DeepSeek-7B instead of zero-shot GPT-4oText2Arch · Text2Arch manual annotation set (99 images)ROUGE-L: 28.2 -> 55.2Improved
use fine-tuned DeepSeek-7B instead of DiagramAgentText2Arch · Text2Arch test setROUGE-L: 42.2 -> 46.8Improved
use fine-tuned DeepSeek-7B instead of zero-shot GPT-4o by GPT judgeText2Arch · GPT-4o judged compatibility (0-5)compatibility score: 2.72 vs 2.68No clear effect
add GPT-4o refinement on top of detector-based DOT2 to form DOT3 (ablation)DOT label pipeline · DOT label quality (manual set)Edge F1: 8.3 -> 51.7Improved
add detector/OCR structure grounding to GPT extraction (DOT3 vs DOT1) (ablation)DOT label pipeline · DOT label quality (manual set)Node F1: 67.5 -> 74.5Improved
use description Desc3 (retrieved paragraphs plus GPT revision) instead of Desc1 or Desc2 (ablation)description pipeline · GPT pairwise preferencepreference rate: over 90%Improved
train arch classifier with CLIP at learning rate 1e-5 instead of 5e-4 (ablation)CLIP classifier · Arch vs no-arch testtest accuracy: 0.5533 -> 0.8345Improved

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