RECORD
Token-Budgeted Candidate Retrieval for 3D Scene-Grounded VLMs: Category Names vs. Pairwise Geometry
Computer Vision and Pattern Recognition · Submitted 9 Oct 2026
abstract compiled · 2026-10-14 · no language-model review
The authors' main claim On ScanNet indoor referring benchmarks (Dai et al., 2017), matching the query to object category names outperforms pairwise geometry scoring at every reported K.
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
Under a fixed context budget, which objects and relations should a 3D scene-grounded VLM see first? On ScanNet indoor referring benchmarks (Dai et al., 2017), matching the query to object category names outperforms pairwise geometry scoring at every reported K. The gaps are largest at high recall. Analytic pairwise cues beat per-object cue scoring and query-agnostic kNN graphs, including on paraphrases. MiniLM and BGE-small still score higher on those splits. We evaluate token-budgeted candidate retrieval with axis-aligned boxes, a token-ratio metric, training-free baselines, two small encoders that match q to each object's category name, and QCRR (Query-Conditioned Relation Retrieval), a deterministic pairwise-geometry scorer. Across ScanRefer gt_val plus a custom 20% scan-id split of Nr3D and Sr3D (35,504 queries; not official ReferIt3D val), additive noun matching reaches R@10 0.864 on ScanRefer against QCRR's 0.626; MiniLM and BGE-small reach 0.911 to 0.913 on ScanRefer and 0.941 to 0.947 on Nr3D. A corrected 3DGraphLLM harness (native Acc@0.25 0.67 on the full graph, n=300 ScanRefer queries) shows that context selection changes downstream accuracy under a matched object budget (Random 0.44, QCRR 0.56, MiniLM 0.67, oracle 0.70). Semantic retrieval beats hand-designed geometry on that harness. We will release the protocol, retriever implementations, and per-query outputs so new selectors can be swapped in without rewriting the analysis.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | Under a fixed context budget, which objects and relations should a 3D scene-grounded VLM see first? | background 0.74 | 0.04 | 0.02 | unclear 1.00 |
| s1 | On ScanNet indoor referring benchmarks (Dai et al., 2017), matching the query to object category names outperforms pairwise geometry scoring at every reported K. | result 1.00 | 0.78 | 0.62 | asserted 1.00 |
| s2 | The gaps are largest at high recall. | result 0.99 | 0.81 | 0.07 | asserted 0.99 |
| s3 | Analytic pairwise cues beat per-object cue scoring and query-agnostic kNN graphs, including on paraphrases. | result 0.97 | 0.62 | 0.12 | asserted 1.00 |
| s4 | MiniLM and BGE-small still score higher on those splits. | result 0.96 | 0.62 | 0.05 | asserted 1.00 |
| s5 | We evaluate token-budgeted candidate retrieval with axis-aligned boxes, a token-ratio metric, training-free baselines, two small encoders that match q to each object's category name, and QCRR (Query-Conditioned Relation Retrieval), a deterministic pairwise-geometry scorer. | method 0.97 | 0.87 | 0.05 | asserted 0.58 |
| s6 | Across ScanRefer gt_val plus a custom 20% scan-id split of Nr3D and Sr3D (35,504 queries; not official ReferIt3D val), additive noun matching reaches R@10 0.864 on ScanRefer against QCRR's 0.626; MiniLM and BGE-small reach 0.911 to 0.913 on ScanRefer and 0.941 to 0.947 on Nr3D. | result 1.00 | 0.88 | 0.24 | asserted 1.00 |
| s7 | A corrected 3DGraphLLM harness (native Acc@0.25 0.67 on the full graph, n=300 ScanRefer queries) shows that context selection changes downstream accuracy under a matched object budget (Random 0.44, QCRR 0.56, MiniLM 0.67, oracle 0.70). | result 0.94 | 0.87 | 0.13 | asserted 1.00 |
| s8 | Semantic retrieval beats hand-designed geometry on that harness. | result 0.82 | 0.87 | 0.29 | asserted 1.00 |
| s9 | We will release the protocol, retriever implementations, and per-query outputs so new selectors can be swapped in without rewriting the analysis. | future work 0.57 | 0.47 | 0.03 | asserted 0.71 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original abstract
Under a fixed context budget, which objects and relations should a 3D scene-grounded VLM see first? On ScanNet indoor referring benchmarks (Dai et al., 2017), matching the query to object category names outperforms pairwise geometry scoring at every reported K. The gaps are largest at high recall. Analytic pairwise cues beat per-object cue scoring and query-agnostic kNN graphs, including on paraphrases. MiniLM and BGE-small still score higher on those splits. We evaluate token-budgeted candidate retrieval with axis-aligned boxes, a token-ratio metric, training-free baselines, two small encoders that match q to each object's category name, and QCRR (Query-Conditioned Relation Retrieval), a deterministic pairwise-geometry scorer. Across ScanRefer gt_val plus a custom 20% scan-id split of Nr3D and Sr3D (35,504 queries; not official ReferIt3D val), additive noun matching reaches R@10 0.864 on ScanRefer against QCRR's 0.626; MiniLM and BGE-small reach 0.911 to 0.913 on ScanRefer and 0.941 to 0.947 on Nr3D. A corrected 3DGraphLLM harness (native Acc@0.25 0.67 on the full graph, n=300 ScanRefer queries) shows that context selection changes downstream accuracy under a matched object budget (Random 0.44, QCRR 0.56, MiniLM 0.67, oracle 0.70). Semantic retrieval beats hand-designed geometry on that harness. We will release the protocol, retriever implementations, and per-query outputs so new selectors can be swapped in without rewriting the analysis.
Human review
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.