Grafting is proposed, a simple manual intervention that exploits knowledge already encoded in the model to improve AFA performance without requiring ground-truth labels at inference time, and extensive evaluations demonstrate that grafting consistently outperforms direct-prompting approaches.
Wendy Zheng, Liang Wu, Yin-Han He et al.· 0 citations
LLM compression reduces inference costs and memory requirements, but selecting a method and configuration remains largely empirical because comparable resource reductions can produce different capability losses. We systematically investigate capability scaling-down laws for LLM compression across pruning, quantization,...
Xue-Qi Cheng, Liang Wu, Kelly Wan et al.· 0 citations
The Code-with-Image Bench (CwI-Bench), thirty task families induced by hidden visual computations with disjoint learning and evaluation splits, even GPT-5.6-luna stays below 30% with tool-free chain of thought; given a bare interpreter it reaches 43%, and with skills evolved through its own executable reflection, 67%.
Tianze Yang, Liang Wu, Rui Sun et al.· arXiv.org· 2 citations
AnyJev is described, which reads a typed decision from one prefill of a pretrained instruction-tuned language model, which divides out a label prior estimated from unlabelled inputs, and it averages log-probabilities over the K cyclic rotations of the option list.
Jia-Mu Zhang, Tian-Ze Yang, Yu-Cheng Shi et al.· 6 citations
It is found that compression degrades reliability before solvability and proposed PAIR (Prompt Adaptation using Interventional Rollouts) for adapting structured compression prompts achieves the strongest cross-run reliability among compressed methods in every main benchmark-scope combination.
Guang-Hui Min, Liang Wu, Ming-Jia Shi et al.· 0 citations
SPECTRA is developed, a training-free, drop-in codec that re-encodes the cache into a coordinate system and concentrates the bit budget on the channels that carry the signal, pushing usable compression past the 2-bit cliff so the same GPU holds longer contexts and larger batches.
Jiamu Zhang, Liang Wu, Kelly Wan et al.· 1 citation
TRACE is introduced, a verifier-guided framework that evaluates individual compaction events through paired closed-loop continuations from the same environment state and uses summary preferences to optimize a natural-language compression prompt while keeping all models frozen.
The results show that skill representation itself matters, and that simply preserving the structure already present in skill files can substantially improve retrieval.
Paimon Goulart, Liang Wu, Ke Wan et al.· 0 citations
MV-WSA (Marginal-Value Working-Set Allocation), which splits memory by marginal latency benefit per byte while enforcing a KV-admission floor, is implemented in WiSP (Working-Set Paging), a routing-aware expert pager that plugs into an unmodified serving engine and preserves byte-identical outputs.
A unified semantic modeling framework powered by a small language model (SLM) to address the challenges of job understanding in structured and unstructured contexts and provides practical insights into building industry-scale text understanding systems.
Daniel Xu, Baofeng Zheng, Jianqiang Shen et al.· Annual International ACM SIG...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.