We have 9 of 44 papers
We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.
Not the right person? Other researchers publish under this name.
On-policy distillation (OPD) supervises student-generated trajectories with token-level teacher signals. Its sampled-token variant avoids the cost of full-vocabulary probabilities. Yet we find that training on fewer tokens can outperform full-token OPD, challenging the intuition that more supervision improves learning....
Speculative decoding accelerates autoregressive generation by using a lightweight draft model to propose multiple tokens that are verified by a target model in parallel. However, the standard acceptance rule focuses on exact distribution correction and rejects tokens that remain highly plausible under the target model...
Multimodal large language models (MLLMs) incur substantial inference costs when processing long visual-textual sequences. While existing operation compression methods exploit modality-level redundancy, they largely treat computation within attention heads and shared feed-forward network (FFN) channels as unified units,...
Extensive experiments across a diverse suite of tasks, including mathematical reasoning, code completion, open-domain QA, and sparse retrieval, demonstrate that Adaptive Mass-Segmented KV Compression consistently mitigates structural fragmentation and boosts model performance.
A three-stage calibration framework that evaluates confidence before, during, and after chain-of-thought generation, corresponding to difficulty estimation, early termination, and answer aggregation finds that OPD provides the most useful pre-reasoning confidence, SFT gives the strongest online signal for early stoppin...
This work recast CoT compression along three dimensions: importance criterion, restructuring level, and compression budget, and yields condition-aware guidelines for matching compression to deployment context.
Results show that separating reusable schema encoding from selective resource access substantially reduces agentic inference costs with limited effectiveness loss.
A three-stage calibration framework that evaluates confidence before, during, and after chain-of-thought generation, corresponding to difficulty estimation, early termination, and answer aggregation finds that OPD provides the most useful pre-reasoning confidence, SFT gives the strongest online signal for early stoppin...
WIDE is presented, the first end-to-end differentiable token-level dynamic width pruning framework designed for both prefill and decode scenarios, and a pruning--kernel co-design framework that decomposes dynamic sparsity acceleration into mask reordering, hardware-agnostic block-level skipping, and hardware-dependent...
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.
