AI Paper Atlas

Research map / Benchmarks / CIRR

CIRR leaderboard

42 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on CIRR. Best R@1: 55.52 (ReCALL (Rrefine), CVPR 2026).

Numbers are as each paper reports them, read by AI from its results tables and not checked by hand. Papers can differ in setting (training data, model size, extra data); the topic column shows what each paper is about. Open the paper before you compare or quote a number.

CIRR · R@1

41 papers, each with its best reported number for its own method; higher is better.

#R@1 ↑MethodPaperTopic
155.52ReCALL (Rrefine)
Qwen2.5-VL-7B
ReCALL: Recalibrating Capability Degradation for MLLM-based Composed Image Retrieval
CVPR 2026 · Table 1
Human and embodied modeling
255.24ConText-CIR ViT-H
OpenCLIP ViT-H
ConText-CIR: Learning from Concepts in Text for Composed Image Retrieval
CVPR 2025 · Table 1
Text-to-image diffusion
354.65MAPNet
BLIP-2
Multi-Schema Proximity Network for Composed Image Retrieval
ICCV 2025 · Table 1
Human and embodied modeling
452.34ReTrackReTrack: Evidence-Driven Dual-Stream Directional Anchor Calibration Network for Composed Video Retrieval
AAAI 2026 · Table 2
Video-language understanding
552.22QuRe
BLIP-2
QuRe: Query-Relevant Retrieval through Hard Negative Sampling in Composed Image Retrieval
ICML 2025 · CIRR performance comparison ta
Autoregressive and latent generation
651.96SPRC
BLIP-2
Sentence-level Prompts Benefit Composed Image Retrieval
ICLR 2024 · Table 2
Human and embodied modeling
751.09HUGHeterogeneous Uncertainty-Guided Composed Image Retrieval with Fine-Grained Probabilistic Learning
AAAI 2026 · Table 2
Human and embodied modeling
850.96BLIP†tdm + NLVR2′se
BLIP ViT-L/16
Visual Delta Generator with Large Multi-modal Models for Semi-supervised Composed Image Retrieval
CVPR 2024 · Table 2
Human and embodied modeling
949.69CoVR-BLIP (Ours), finetuned
BLIP ViT-L
CoVR: Learning Composed Video Retrieval from Web Video Captions
AAAI 2024 · Table 3
Video-language understanding
1047.24DIPNEC
CLIP ViT-L/14
Aligning Composed Query with Image via Discriminative Perception from Negative Correspondences
AAAI 2025 · Table 2
Human and embodied modeling
31 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

CIRR · R@5

10 papers, each with its best reported number for its own method; higher is better.

#R@5 ↑MethodPaperTopic
182.53ReTrackReTrack: Evidence-Driven Dual-Stream Directional Anchor Calibration Network for Composed Video Retrieval
AAAI 2026 · Table 2
Video-language understanding
280.2DIPNEC
CLIP ViT-L/14
Aligning Composed Query with Image via Discriminative Perception from Negative Correspondences
AAAI 2025 · Table 4
Human and embodied modeling
375.41LIMN+Self-Training Boosted Multi-Factor Matching Network for Composed Image Retrieval
TPAMI 2023 · Table 2
Human and embodied modeling
466.7CoVR-BLIP (Ours)
BLIP ViT-L
CoVR: Learning Composed Video Retrieval from Web Video Captions
AAAI 2024 · Table 3
Video-language understanding
565.35Proposed approach
CLIP RN50x4
Effective Conditioned and Composed Image Retrieval Combining CLIP-Based Features
CVPR 2022 · Table 2
Autoregressive and latent generation
656.84MCL
Llama2-7B
Improving Context Understanding in Multimodal Large Language Models via Multimodal Composition Learning
ICML 2024 · Table 1
Multimodal LLMs
756.5FlowCIR
CLIP ViT-B/32
FlowCIR: Semantic Transport via Flow Matching for Zero-Shot Composed Image Retrieval
ECCV 2026 · Table 1
CLIP and zero-shot
854.8KEDs
CLIP ViT-L/14
Knowledge-Enhanced Dual-stream Zero-shot Composed Image Retrieval
CVPR 2024 · Table 4
CLIP and zero-shot
951.7Pic2Word
CLIP ViT-L/14
Pic2Word: Mapping Pictures to Words for Zero-Shot Composed Image Retrieval
CVPR 2023 · Table (CIRR)
Zero-/few-shot and meta-learning
1046.1ARTEMIS (BiGRU)
ResNet152 features + BiGRU
ARTEMIS: Attention-based Retrieval with Text-Explicit Matching and Implicit Similarity
ICLR 2022 · CIRR dataset, test set table
Text-to-image diffusion

Not affiliated with any conference, journal or the benchmark's authors. Corrections: contact@latenticlabs.com.