Tail subtraction is introduced, which removes shared prompt and continuation semantics from boundary states and yields cleaner, more stable steering signals, and suggests that steering depends on representations of what the model is about to do, not merely on what has already appeared.
Jiaran Ye, Lingxu Ran, Zijun Yao et al.· arXiv.org· 2 citations
Results indicate that scalable LGN depth requires both stable optimization and credit-preserving information access, and introduce Input-Anchored Logic Gate Networks (IALGN), in which each gate combines a private hidden spine with a direct input anchor.
Taegun An, Dohun Kim, Haebeom Lee et al.· 0 citations
GHR-VLM, a visual grounded hybrid reasoning framework for zero-shot transit-bus video analytics, is proposed, motivated by the observation that explicit visual grounding can improve VLM reasoning by converting long surveillance streams into compact, passenger-centered spatiotemporal evidence.
Kaicong Huang, Weiheng Oh, Jack M. Reilly et al.· 0 citations
This work investigates the ability of Large Language Models to generate structurally valid and constraint-compliant network topologies through a constraint-driven pipeline combining hierarchical modeling and systematic validation, and provides a systematic benchmark for understanding how LLMs handle structural and resilience constraints in topology synthesis.
This work borrows definitions from narratology to analyze eight intricate dimensions of character, such as stylization and wholeness, which consider more than just basic characteristics of characters within LLM and human-written stories.
A. Brei, Abhisheik Sharma, Nicholas Sanaie et al.· Annual Meeting of the Associ...· 0 citations
The transformer reduces multiplication to addition in discrete-log space, implementing a "Discrete-Log Clock" algorithm analogous to Nanda et al.'s Clock algorithm for addition, which generalizes: matching the analysis basis to the algebraic structure of the task reveals interpretable structure where standard tools see noise.
TokenPilot is presented, a dual-granularity context management framework that reduces costs by 61% and 56% in isolated mode, and 61% and 87% in continuous mode, while maintaining competitive performance compared to prior systems.
Buqiang Xu, Z. Xue, Dian Chen et al.· arXiv.org· 1 citation
Sycophancy co-occurs with degraded judged truthfulness (rho=0.40), a coupling that strengthens across generations, and a single direct instruction outperforms an elaborate reasoning protocol in seven of eight variants.
DiffuSent is presented, a non-auto-regressive diffusion framework that systematically formulates all ABSA subtasks as boundary denoising diffusion processes, progressively refining boundaries over noisy states, and introduces a contrastive denoising training strategy which effectively address duplicate predictions with subtle variations introduced by diffusion process.
S. Long, Yanglei Gan, Xuchuan Zhou· arXiv.org· 0 citations
LongDS is introduced, a benchmark for long-horizon, multi-turn data analysis where agents must maintain, update, restore, and compose evolving analytical states, suggesting that the key bottleneck is maintaining a correct analytical state rather than increasing interaction budget.
This work presents a systematic study of scale vectors in LLMs from the perspectives of expressivity, optimization, and architectural structure, and proposes three lightweight and complementary improvements to scale vectors: branch-specific heterogeneity, improved placement around linear mappings, and magnitude-direction reparameterization.
This work proposes Mixture of Activations (MoA), a token-adaptive FFN design that mixes a dictionary of activation functions using lightweight input-dependent gates while sharing the same linear projections, suggesting that token-adaptive activation mixing is a simple and effective mechanism for improving FFN expressivity in LLMs.
