Interactive retrieval with partial evidence constitutes a sequential information-acquisition problem: an agent must choose questions that acquire useful evidence for the next retrieval update. Existing systems train this decision by imitating an offline ordering of candidate QA pairs. However, a question's value depends on the response it elicits and its downstream effect on retrieval. We establish that candidate discriminativeness and perceived usefulness provide weak supervision for this objective, then introduce RAVEL, a retrieval-aware online reinforcement learning framework for interactive person re-identification. RAVEL initializes from supervised question generation, observes the current Top-4 candidates directly, and optimizes the question policy with rank feedback from the full question-answer-retrieval loop. Experiments on Interactive-PEDES show that RAVEL delivers progressively stronger retrieval performance across five interaction rounds. Further analysis shows that RAVEL allocates more of its interaction budget to localized open-ended prompts targeting specific attributes.
Figures & tables
Figure 1: Illustrative interactive ReID example. From the same Top-4 state at Rank 31, the jacket question moves the target to Rank 38, whereas the retrieval-aware shoe question raises it to Rank 2; question value follows the evidence elicited and its effect on re-ranking.
Figure 2: Question-quality judgments and downstream retrieval gain, measured by the reciprocal-rank difference ΔRR=1/rafter−1/rbefore .
Figure 3: Illustration of layer-wise visual attention used to compute visual reliance.
Figure 4: Overview of RAVEL’s three-stage training pipeline.
Method
Round 3
Round 5
BRI ↓
R@1
R@5
R@10
mAP
R@1
R@5
R@10
mAP
Initial
37.61
61.74
72.01
28.23
37.61
61.74
72.01
28.23
-
PlugIR ( Lee et al., 2024 )
43.15
67.17
76.78
30.92
47.15
70.83
79.73
33.80
1.012
ChatIR ( Levy et al., 2023 )
45.02
69.24
78.60
33.07
49.06
72.94
81.57
35.85
0.997
GPT-5.6 Luna ( OpenAI, 2026 )
46.37
67.73
77.54
35.43
47.88
69.27
78.19
36.56
1.042
SimRV ( Liang and Albanie, 2023 )
62.91
83.26
89.18
38.65
63.42
84.21
89.34
39.46
0.720
Table 1: Interactive retrieval performance on Interactive-PEDES.
Figure 5: Analysis of learned questioning behavior and retrieval-text quality. (a) Ratio of local WH/open questions over interaction rounds. (b) Mean retrieval gain of different question types. (c) Quality of the final retrieval text, measured by query length, negative-word count, and attribute coverage.
Method
CUHK-PEDES
ICFG-PEDES
RSTPReid
R@1
R@5
R@10
mAP
R@1
R@5
R@10
mAP
R@1
R@5
R@10
mAP
CFine ( Yan et al., 2023 )
69.57
85.93
91.15
-
60.83
76.55
82.42
-
50.55
72.50
81.60
-
RaSa ( Bai et al., 2023 )
76.51
76.51
94.25
69.38
65.28
80.40
85.12
41.29
66.90
86.50
91.35
52.31
APTM ( Yang et al., 2023 )
76.53
90.04
94.15
66.91
68.51
82.99
87.56
41.22
67.50
85.70
91.45
52.56
AUL ( Li et al., 2024b )
77.23
90.43
94.41
-
69.16
83.32
88.37
-
71.65
87.55
92.05
-
RDE ( Qin et al., 2024 )
76.20
90.53
94.20
67.85
67.83
82.50
87.28
40.79
67.40
85.65
90.60
52.01
Table 2: Comparison of text-based ReID retrievers and interactive questioners on three benchmarks.
Table 8
Table 7: Prompt used by the Qwen3-VL question-quality evaluator.
Table 10
Figure 6: Attribute shares before and after retrieval-aware training.
Figure 7: Qualitative comparison on a case where the target starts at Rank 5.
Figure 8: Qualitative comparison on a case where the target starts at Rank 52.
Person re-identification (ReID) ranks an image gallery using either an image query or a textual query. In image-to-image (I2I) ReID, viewpoint, illumination, occlusion, and background changes can make distractors appear more similar than the true match. In text-to-image (T2I) ReID, incomplete or ambiguous descriptions may omit key visual evidence. We study this problem for vision-language model (VLM) based ReID, where attributes such as clothing color and carried objects provide useful fine-grained matching cues. However, directly using attribute-prompt similarities as retrieval scores is unreliable because templated prompts can be biased, correlated, and poorly calibrated. We therefore propose PRompt-guided Attribute-based Inference-time Similarity Enhancement (PRAISE), a query-side test-time adaptation framework for VLM-based ReID. Instead of treating prompts as scores, PRAISE uses a transferable attribute-prompt bank to construct an Attribute-Conditioned Subspace (ACS): a query-dependent low-rank subspace induced by the most compatible attribute prompts. ACS provides a controlled attribute-conditioned geometry for reshaping query--gallery affinities and guiding lightweight inference-time updates without identity or attribute labels. Experiments across multiple I2I- and T2I-ReID benchmarks show that PRAISE consistently improves VLM retrieval backbones and outperforms both generic and ReID-specific TTA baselines.
Learning identity-discriminative representations with multi-scene generality has become a critical objective in person re-identification (ReID). However, mainstream perception-driven paradigms tend to identify fitting from massive annotated data rather than identity-causal cues understanding, which presents a fragile representation against multiple disruptions. In this work, ReID-R is proposed as a novel reasoning-driven paradigm that achieves explicit identity understanding and reasoning by incorporating chain-of-thought into the ReID pipeline. Specifically, ReID-R consists of a two-stage contribution: (i) Discriminative reasoning warm-up, where a model is trained in a CoT label-free manner to acquire identity-aware feature understanding; and (ii) Efficient reinforcement learning, which proposes a non-trivial sampling to construct scene-generalizable data. On this basis, ReID-R leverages high-quality reward signals to guide the model toward focusing on ID-related cues, achieving accurate reasoning and correct responses. Extensive experiments on multiple ReID benchmarks demonstrate that ReID-R achieves competitive identity discrimination as superior methods using only 14.3K non-trivial data (20.9% of the existing data scale). Furthermore, benefit from inherent reasoning, ReID-R can provide high-quality interpretation for results.
Person re-identification (ReID) is essential for multi-camera surveillance and tracking, yet remains difficult due to viewpoint and illumination changes, occlusion, background clutter, and low resolution imagery. We propose a generative AI integrated multimodal ReID framework designed explicitly for robustness under missing cues and low latency deployment. The key idea is a cost aware early-exit cascade that prioritizes inexpensive, high confidence evidence and only triggers expensive modalities for ambiguous cases. Our system integrates (i) global visual embeddings from segmented person regions, (ii) automatically generated fine grained semantic attribute descriptions generated by vision-language models (VLMs), and (iii) optional facial embeddings when face observations are reliable. To optimize the balance between accuracy and latency, we use a cost aware early-exit cascade instead of fusing all modalities. Specifically, we first inspect the top-k retrieval results to determine whether the query is unambiguous. If the best match is clearly separated from the remaining candidates, we stop early and return the result to minimize latency; in ambiguous cases, we keep multiple hypotheses and invoke additional modalities (face/semantic) with adaptive reliability weighting to refine the decision. We report person re-identification performance using mAP and Rank-1 accuracy on the Market-1501 and DukeMTMC-reID benchmarks. The proposed adaptive early-exit cascade resolves 60.7% of DukeMTMC-reID queries and 68.4% of Market-1501 queries without invoking semantic reasoning, reducing computational overhead while maintaining competitive retrieval performance.
Leon Fernando, C Dombawala, P. Hettigoda +3
University of Moratuwa · Zone24x7 (Pvt) Ltd · Chulalongkorn University +2