RenderRank: Learning to Rerank Text with Compressed Visual Tokens
RenderRank reranks documents from compressed visual tokens, using 16.5-35.5% fewer tokens and beating sub-4B text baselines.
RenderRank is a document reranker that scores relevance from compressed visual tokens rather than text token sequences. Training first matches a text-based teacher, then refines relative scores of positive and negative documents for each query. On 11 BEIR datasets it uses 16.5-35.5% fewer input tokens and reaches average NDCG@10 of 55.96, beating evaluated text baselines under 4B parameters and some larger models. On four long-document datasets it averages NDCG@10 of 88.27 with about half the tokens and 1.70x the best baseline throughput.
- Scores query relevance from compressed visual document tokens instead of text sequences.
- Training aligns a text teacher, then refines positive versus negative document scores.
- Average BEIR NDCG@10 is 55.96 with 16.5-35.5% fewer input tokens.
- Long-document NDCG@10 averages 88.27 at about half the tokens and 1.70x throughput.
Full article194 words · extracted from huggingface.co · click to collapse
Rendering document text as images allows vision-language models to encode documents as visual tokens, which can reduce input sequence length compared with text input. This reduction in input length is particularly useful for reranking, where each query involves scoring multiple candidate documents and token savings apply to each candidate evaluation. We introduce RenderRank, a reranker that learns query-dependent relevance scoring from compressed visual document representations instead of the text token sequences used by conventional text-based rerankers. Training first aligns relevance scores from visual inputs with those of a text-based teacher, then refines the relative scores of positive and negative documents for the same query. Across 11 datasets from BEIR, RenderRank uses 16.5-35.5% fewer input tokens while achieving an average NDCG@10 of 55.96, outperforming all evaluated text-based baselines below 4B parameters and some larger models. Across four long-document datasets, it achieves an average NDCG@10 of 88.27 with approximately half the average input token count of the evaluated text-based rerankers. In this setting, RenderRank delivers 1.70x the highest average throughput of the evaluated baselines. These results demonstrate that compressed visual representations can support accurate document relevance scoring, providing an alternative to text token representations for reranking.
Text extracted automatically; images, tables and formatting may be missing. Original: https://huggingface.co/papers/2609.35069