AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

LIRA: Reasoning Reconstruction via Multimodal Large Language Models

Zhen Zhou, Tong Wang, Yunkai Ma, Xiao Tan, Fengshui Jing

ICCV 2025 · Vision-language and multimodal › Multimodal LLMs · approach: Spatial reasoning in VLMs · 1 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 confidence branch for candidate masks (ablation)LIRA · ReasonRecon testAP50: 29.30 -> 34.39Improved
use fine-tuned LLaVA-style reasoning segmentation module (final) instead of SEEM (ablation)LIRA · ReasonRecon testAP50: 11.00 -> 34.39Improved
use own 2D reasoning module instead of fine-tuned LISA (ablation)LIRA · ReasonRecon testAP50: 22.66 -> 34.39Improved
use own 2D reasoning module instead of fine-tuned LLaVA-Grounding (ablation)LIRA · ReasonRecon testAP50: 34.26 -> 34.39No clear effect
add text features to instance fusion (ablation)LIRA · ReasonRecon testAP50: 33.13 -> 34.39Improved
add bbox features to instance fusion (ablation)LIRA · ReasonRecon testAP50: 33.21 -> 34.39Improved
add voxel features to instance fusion (ablation)LIRA · ReasonRecon testAP50: 33.68 -> 34.39Improved
fuse multiple keyframes simultaneously within FBV (ablation)LIRA · ReasonRecon testAP50: 34.15 -> 34.39Improved
use ChatGPT-4o-mini instead of Llama3 in stage III (ablation)LIRA · ReasonRecon testAP50: 33.27 -> 34.39Improved
use ChatGPT-4o-mini instead of Qwen2.5 in stage III (ablation)LIRA · ReasonRecon testAP50: 33.58 -> 34.39Improved
use ChatGPT-4o instead of ChatGPT-4o-mini in stage III (ablation)LIRA · ReasonRecon testAP50: 34.39 -> 34.74No clear effect
use larger MLLM (LLaVA-13B) and ChatGPT-4o for reasoningLIRA · ReasonRecon testAP50: 34.39 -> 35.32Improved

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