2026· International Conference on Principles and Practice of Constraint Programming· pp. 18:1-18:18· 0 citations· 40 references
Computer Science
TL;DR
This paper proposes combining a Constraint Programming model with a Masked Language Model ( mlm) to perform Large Neighbourhood Search ( lns), and shows that it can quickly generate many high-quality sentences and molecules, even for highly-constrained tasks.
This work proposes a mechanism that bridges these two paradigms: an LLM produces a high-quality seed architecture, then decomposes it into a "slotted architecture", a scaffold with named, interchangeable module slots that automatically defines a bounded, task-specific search space for conventional NAS to explore, witho...
Seokhoon Jeong, Mijung Kim, Taehwan Kim· 0 citations
This paper argues that the key lies in designing an appropriate search-state representation that connects the internal knowledge of LLMs with the structural requirements and constraints of visual-token pruning, and proposes AutoPrune, a training-free framework for LLM-driven visual-token pruning policy design.
Zhen Liu, Wenli Huang, Wei Song et al.· 0 citations
The results show how resource-rational models of WM allocation can be implemented in neural models simply and successfully, and point to a dissociation between WM retrieval mechanisms and the underlying memory representations in models of human sentence processing.
Weijie Xu, Brian Dillon, Richard Futrell· Annual Meeting of the Associ...· 1 citation
A novel and interpretable approach to perform symbolic integration using deep learning through integral rule prediction to speed up the search and introduces the first-of-its-kind symbolic integration rules dataset comprising two million distinct functions and integration rule pairs.
It is observed that attention scores from both vision and text tokens peak at modality separator tokens, suggesting that these separators bridge the two modalities and proposes SepPrune, an efficient, training-free, plug-and-play pruning method that uses the separator token as a unified query to rank and select informa...
Yucheng Wang, Qihui Zhu, Yang Liu et al.· arXiv.org· 0 citations
A new training and inference algorithm inspired by parallel-tempering Markov chain Monte Carlo methods is developed, which is named parallel noising and allows NMLNs to match the performance of specialized text-based recurrent models when generating small molecular structures.
Peter Jung, Giuseppe Marra, Ondřej Kuželka· arXiv.org· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
