[
    {
        "id": "osp-21630",
        "type": "article-journal",
        "title": "Recovering Off-Policy Supervision for Speculative Decoding",
        "author": [
            {
                "family": "Lee",
                "given": "Jungseob"
            },
            {
                "family": "Park",
                "given": "Chanjun"
            },
            {
                "family": "Eo",
                "given": "Sugyeong"
            },
            {
                "family": "Moon",
                "given": "Hyeonseok"
            }
        ],
        "URL": "https://omanscience.com/en/articles/recovering-off-policy-supervision-for-speculative-decoding",
        "language": "en",
        "issued": {
            "date-parts": [
                [
                    2026
                ]
            ]
        },
        "abstract": "Block drafters for speculative decoding are commonly trained on corpora written by external models, where a single off-policy token invalidates supervision for all subsequent slots in a block. Existing approaches discard these divergent slots, resulting in severe supervision loss. To resolve this problem while preserving the training corpus, we propose a rollout-based training framework that recovers full supervision through two complementary components. The first component, Anchor-Label Relabelling (ALR), replaces corpus labels with distributions from greedy target rollouts, restoring valid supervision across all predicted slots. The second component, In-Rollout Anchors (IRA), places draft blocks directly inside these rollouts to expose the drafter to target-generated context, reusing precomputed rollout features at no additional target cost. Across fixed vision-language and text corpora, our framework increases greedy accepted length by up to 36.5% over DFlash and consistently outperforms erasing baselines. Notably, a single epoch of our method surpasses the best erase schedules. After three epochs, it matches the acceptance length of training on target-regenerated responses. These results show that our framework provides an effective and compute-efficient approach for training speculative drafters on fixed corpora without modifying the original text. Code is available at https://github.com/js-lee-AI/ALR-IRA."
    }
]