@inproceedings{ting-etal-2026-bridging,
    title = "Bridging the Memorization-Utilization Gap: Near-Lossless Context Compression via Reinforcement Learning",
    author = "Ting, Yujan  and
      Tang, Xu  and
      Chen, Terrence  and
      Huang, Weijing",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.acl-long.682/",
    pages = "14949--14972",
    ISBN = "979-8-89176-390-6",
    abstract = "Despite recent progress in context compression, we identify a fundamental memorization-utilization gap where models can compress context with near-perfect fidelity yet fail to effectively utilize these compressed representations for downstream tasks. We address this with a holistic training paradigm spanning pretraining, instruction tuning, and reinforcement learning, built upon an average pooling compression. Our key innovation uses outcome-based RL to enable implicit expansion: the model learns to adaptively unfold task-relevant details during generation, interleaving reconstruction with reasoning. We achieve near-lossless 16x context compression ({\ensuremath{\approx}}5.3x decoder sequence-length reduction in our current implementation) across 7B and 32B models, recovering over 98{\%} of full-context QA performance and outperforming prior methods by 11 points. Our 32B model demonstrates strong out-of-distribution and length generalization, robustly scaling to 120k-token contexts despite training on no more than 4k tokens, matching full-context performance on NIAH, LongBench v2, and multi-hop reasoning. We verify the implicit expansion behavior in experiments."
}