mlsanityData to 14 Oct 2026

G2D: Generative-to-Discriminative Collaborative Inference for Zero-Shot Image Classification

arXiv 2608.26744 ↗First posted 27 Aug 2026cs.CVcited by 0

On the map: LLM reasoning → Multimodal reasoning

also in 1 more
Abstract (arXiv)

Zero-shot classification needs efficient label retrieval and fine-grained visual reasoning, yet discriminative and generative vision-language models fail in complementary ways.When CLIP's top-1 prediction is wrong, the correct label often remains in its top-K shortlist, making disambiguation rather than recall the key challenge.Standalone generative models, however, are hindered by large label spaces and unconstrained outputs.This complementarity motivates separating broad candidate retrieval from fine-grained, image-grounded verification.We propose G2D, a training-free framework that uses a generative VLM to verify CLIP-retrieved candidates against the image.Candidate names and CLIP probabilities provide a structured prior for resolving visually similar classes.Fixed confidence routing, entropy-adaptive candidate sizing, and trie-constrained decoding focus generative reasoning on uncertain samples and ensure one valid output for each input at test time.Across eight benchmarks, G2D achieves 68.85% average accuracy, versus 59.35% for CLIP and 63.11% for the standalone VLM.Across seven generator configurations, candidate-set verification improves average accuracy by 1.08--27.42 percentage points.G2D also transfers to DCLIP, WaffleCLIP, and CuPL, supporting a practical interface between discriminative proposal and generative visual reasoning. Code: https://github.com/Harzva/G2D

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 19 of its 57 references that are in the map. First posted on arXiv 27 Aug 2026.

Builds on

19 of 57 references.

References · 19

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2025Qwen3-VL Technical Reportcited by 2,793
  3. 2013Describing Textures in the Wildcited by 2,125
  4. 2019Do ImageNet Classifiers Generalize to ImageNet?cited by 1,515
  5. 2013Zero-Shot Learning Through Cross-Modal Transfercited by 237
  6. 2016Learning Deep Representations of Fine-grained Visual Descriptionscited by 182
  7. 2014Evaluation of Output Embeddings for Fine-Grained Image Classificationcited by 166
  8. 2023Prompt, Generate, then Cache: Cascade of Foundation Models makes Strong Few-shot Learnerscited by 152
  9. 2022CALIP: Zero-Shot Enhancement of CLIP with Parameter-free Attentioncited by 103
  10. 2023Waffling around for Performance: Visual Classification with Random Words and Broad Conceptscited by 101
  11. 2022K-LITE: Learning Transferable Visual Models with External Knowledgecited by 79
  12. 2024Improved Zero-Shot Classification by Adapting VLMs with Text Descriptionscited by 54
7 more
  1. 2017Fine-graind Image Classification via Combining Vision and Languagecited by 51
  2. 2017Link the head to the "beak": Zero Shot Learning from Noisy Text Description at Part Precisioncited by 46
  3. 2022Improving Zero-shot Generalization and Robustness of Multi-modal Modelscited by 45
  4. 2023Prompting Scientific Names for Zero-Shot Species Recognitioncited by 14
  5. 2019TGG: Transferable Graph Generation for Zero-shot and Few-shot Learningcited by 10
  6. 2024Enhancing Visual Classification using Comparative Descriptorscited by 2
  7. 2026Delving Aleatoric Uncertainty in Medical Image Segmentation via Vision Foundation Modelscited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Will It Zero-Shot?: Predicting Zero-Shot Classification Performance For Arbitrary Queries
  2. 2024Learning by Correction: Efficient Tuning Task for Zero-Shot Generative Vision-Language Reasoning
  3. 2025No Labels Needed: Zero-Shot Image Classification with Collaborative Self-Learning
  4. 2026Explaining CLIP Zero-shot Predictions Through Concepts
  5. 2022VL-Taboo: An Analysis of Attribute-based Zero-shot Capabilities of Vision-Language Models
  6. 2024Image-Caption Encoding for Improving Zero-Shot Generalization
  7. 2024BaFTA: Backprop-Free Test-Time Adaptation For Zero-Shot Vision-Language Models
  8. 2024Seeing the Undefined: Chain-of-Action for Generative Semantic Labels