OpticalRAG compresses retrieved documents to 16 visual tokens with best compression-aware accuracy
kalyan_kpl · x · 2026-11-11
OpticalRAG is a pixel-space compression RAG framework: render documents as page images, encode them with a frozen vision encoder, and distill retrieved visual tokens to as few as 16 injected tokens for the LLM. It works without expensive cross-modal training via two ingredients — Encoder-Bridged Retrieval and Query-Driven Distillation. It achieves the best compression-aware accuracy (CAP) on LongBench-v2, BAMBOO, and LooGLE-v2 while injecting only 16 tokens.
More from Research
- Explosion launches Ellf directory: 52 NLP templates plus multi-agent annotation workflow — _inesmontani · 2026-11-11
- STABLE Optimizer Accepted at NeurIPS 2026: Taming Continual Learning Forgetting at the Optimizer Level — v_lomonaco · 2026-11-11
- From Bahdanau to FlashAttention: a timeline of attention's 8-year evolution — cneuralnetwork · 2026-11-11
- Three years building CPUs and spiking nets: one engineer's path to AI accelerators — k7agar · 2026-11-11
- When AI becomes an oracle, what happens to mathematics? — stevenstrogatz · 2026-11-11
- 200 Years of Linear Algebra in One Chart: From Leibniz's Determinants to Peano's Vector Spaces — burny_tech · 2026-11-11