AI Trends

Trending story

RecSA Outperforms All Baselines in the RULER Long-Context Benchmark

1 reports1 story updatesUpdated Nov 19 03:20

Understand this story

Story summary

The English summary is being updated. See the reporting timeline below.

Latest development · Report

UCSD Hao AI Lab Introduces RecSA, Scaling a Model Trained on 32K Sequences to 2M Tokens at Logarithmic Cost

Developments

1 confirmed developments

  1. English translation is being prepared.

    Source:UCSD Hao AI Lab

Reports and discussion timeline

Browse reports and discussion by time. Their evidence levels remain distinct.

Nov 19

  1. UCSD Hao AI LabFirst-party

    UCSD Hao AI Lab Introduces RecSA, Scaling a Model Trained on 32K Sequences to 2M Tokens at Logarithmic Cost

    UCSD Hao AI Lab introduced RecSA, which builds a tree over the KV cache and searches it layer by layer, making the cost per query grow as O(log L) and enabling a model trained only on 32K sequences to extrapolate to 2M tokens. RecSA was initialized from Qwen3-4B and trained on 0.62B tokens. It outperformed all baselines on RULER at context lengths ranging from 131K to 2M, scoring 60.5 at 1M versus 40.0 for the strongest baseline, Dense + DCA. On a single H200, RecSA achieved a 3.2× prefill speedup and a 3.8× decoding speedup at 524K. The project has open-sourced its code, kernels, Qwen3 integration, and evaluation tools.

Related stories

  1. Related storiesRecSA Achieves Extrapolation to 2M Tokens with Logarithmic Query Cost
  2. Related storiesRecSA Achieves Logarithmic Query Costs with a KV Cache Tree
  3. Related storiesRecSA Extrapolates Context to 2 Million Tokens with Logarithmic Query Cost
  4. Related storiesRecSA Outperforms All Baselines in the RULER Evaluation for Ultra-Long Contexts