SEER通过智能选图压缩，高效处理长文本推理。
SEER: Long-Context Reasoning via Selective Visual-Text Compression
- 按查询相关性选择性压缩文本为图像，仅在需要时还原
- LongBench上平均准确率达51.11%，优于基线2.33~3.49点
- 适合需要长文本精炼推理的场景，如文档问答
长文本推理对大语言模型而言仍计算昂贵，因注意力机制随文本长度呈二次增长。视觉-文本压缩通过将文本转为图像并用视觉语言模型处理，可显著减少标记数量。但现有方法采用统一压缩，忽略查询相关性，可能损失细节精度。我们提出SEER框架，通过视觉扫描学习选择与查询相关的图像，并仅在必要时检索原文内容，结合了视觉压缩的效率与文本推理的精度。基于工具交互轨迹的监督微调使SEER能自适应地选择和检索。在长文本基准测试中，SEER在保持平均提示标记数节省的同时提升了提取精度。在LongBench上，平均准确率为51.11%，比视觉-文本基线Glyph-9B高出2.33点，比Qwen3-8B高3.49点。代码已开源。
原文摘要 · Abstract (English)
Long-context reasoning remains computationally expensive for large language models due to the quadratic complexity of attention over text tokens. Visual-text compression offers a promising alternative by rendering text into images and processing them with vision-language models, often reducing token usage. However, existing approaches apply uniform compression regardless of query relevance, potentially sacrificing precision where detailed extraction is required. We present SEER, a framework that learns to select query-relevant images through visual scanning and retrieve textual content only where needed, combining the efficiency of visual compression with the precision of text-based reasoning. Through supervised fine-tuning on tool-interaction trajectories, SEER learns adaptive tool invocation for selection and retrieval. Experiments on long-context benchmarks show that SEER improves extraction precision through selective text retrieval while retaining average prompt-token savings relative to full-text baselines. On LongBench, SEER achieves 51.11% average accuracy, outperforming the visual-text baseline Glyph-9B by 2.33 points and Qwen3-8B by 3.49 points. Code can be accessed at https://github.com/jiaweixu98/SEER
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。