NeFut Logo NeFut
Admin Login

[CS.AI] From Transient Prompts to Persistent Control: Scientific Poster Generation via Recursive Semantic‑Geometric Contracts

Published at: 2026-09-17 22:00 Last updated: 2026-09-18 00:46
#algorithm #AI #Machine Learning

Scientific poster generation condenses a multimodal paper into a single‑page visual artifact, forcing a strict trade‑off between information coverage and readability within a fixed spatial budget. Existing pipelines treat plans as transient prompts and validate each stage in isolation, which lets requirements drift between content and layout modules and silently invalidates earlier checks. PosterVisor shifts generation from transient prompts to persistent control. An Orchestrator extracts rubrics from the paper and visual assets, compiling them into a Semantic‑Geometric Contract (SGC) that binds claims and sources to required visuals, budget limits, and geometric commitments. Only fully instantiated records become executable assertions; other usable requirements remain soft guidance. Recursive Contract Enforcement (RCE) dynamically triggers cross‑stage checks as evidence emerges. During repairs, RCE re‑evaluates the affected checkpoint states, preventing repair‑induced regressions from propagating silently. PosterVisor is instantiated in HTML/CSS and editable PPTX generators. On the 100‑paper Paper2Poster benchmark, PosterVisor‑PPT raises mean poster‑grounded QA accuracy to 64.47 % (vs. 58.53 %) and is preferred by human judges in 72.5 % of non‑tied pairwise comparisons (95 % CI, 61.6‑83.4 %). A secondary study on 30 papers also yields higher VLM Overall and PaperQuiz means. These findings support rubric‑compiled contracts and stage‑conditioned enforcement for controllable poster synthesis.

Review

Original Source: https://arxiv.org/abs/2609.17326

[h] Back to Home