arXiv:2511. 05650v2 Announce Type: replace-cross Abstract: Alignment has greatly improved large language models (LLMs)' output quality at the cost of diversity, yielding highly similar outputs across generations, especially in open-ended generation tasks.
arXiv:2608.29291v1 Announce Type: new Abstract: Unified multimodal models jointly support understanding and generation, but incur substantial redundant computation across tokens, layers, and generati...
The paper introduces Mahalanobis-Ensemble Decoding (ME-Decoding), a new framework for Large Language Model decoding that treats candidate token selection as an ensemble pruning problem. It uses a Mahalanobis distance-driven objective to promote semantic diversity while maintaining high probabilities, employing a token similarity matrix built with an adaptive-bandwidth kernel over token embeddings. An efficient greedy algorithm with near-linear complexity and theoretical guarantees makes ME-Decoding a plug‑and‑play module with negligible inference overhead, and experiments show strong performance across reasoning and generation tasks.
arXiv:2604. 26170v2 Announce Type: replace Abstract: Adapting large language models (LLMs) to a targeted task efficiently and effectively remains a fundamental challenge.
CAMFT is a Conflict‑Aware Mergeable Fine‑Tuning method designed to make task adaptation efficient and merge‑aware for large language models. Unlike existing approaches that only resolve parameter conflicts after fine‑tuning, CAMFT shapes mergeability during training by guiding each task to update sparse coordinates with lower cross‑task conflict. Experiments show that CAMFT outperforms standard fine‑tuning baselines in multi‑task merging scenarios.
arXiv:2601. 05106v5 Announce Type: replace Abstract: Large language models (LLMs) exhibit strengths across diverse domains.
arXiv:2605.27971v2 Announce Type: replace-cross Abstract: When large language models are fine-tuned to generate persona- or tone-conditioned responses, their output diversity is severely limited--a f...
SwiftExplorer is a training‑free diffusion model alignment plugin that addresses two key issues in objective‑guided sampling: the loss of diversity due to strong directional bias and the inefficiency of constant guidance. It introduces an Inheritance‑Restart exploration mechanism to prevent early convergence and enhance the likelihood of high‑reward trajectories, while a Quality‑Efficiency arbitration mechanism removes incorrect signals and dynamically stops generation when optimal reward gain is achieved. Experiments show that SwiftExplorer improves preference, fidelity, diversity, and richness across multiple evaluation metrics.
Gacha Decoding is an inference-time method that elicits diverse language model generations by treating diversity as an instruction-following problem. It combines a model’s instruction-following ability with randomness from an external RNG to identify and realize distinct response modes, outperforming existing diversity approaches across open-ended domains. The method scales with model capability, achieving higher diversity at equal quality and discovering novel modes with far fewer samples.
arXiv:2607. 20481v1 Announce Type: new Abstract: Local-cloud collaboration is a practical way to deploy large language models under resource constraints, but existing methods often rely on trained routers or collaboration-aware finetuning that tie routing behavior to a particular operating regime.
The paper introduces a new method for aligning cross‑lingual representations in decoder‑only large language models by leveraging the outputs of mixture‑of‑experts (MoE) routers instead of traditional hidden‑state alignment losses. This approach pools router outputs over many tokens, providing more reliable sequence‑level comparisons across languages. Experiments on four open‑source MoEs show that the routing loss not only aligns hidden representations but also improves multilingual performance across a diverse evaluation suite.
The paper introduces Align‑LoRA, a unified LoRA framework for multi‑task learning that replaces complex, isolated adapter designs with a single‑adapter model enhanced by a higher rank and an explicit alignment loss. It demonstrates that a router‑free, multi‑head model with high inter‑head redundancy can outperform more elaborate baselines, and that a unified LoRA can achieve competitive performance while enabling weight merging and zero inference latency. Extensive experiments and theoretical analysis confirm that Align‑LoRA surpasses prevailing approaches, offering a simpler, production‑friendly paradigm for parameter‑efficient fine‑tuning of large language models.
arXiv:2609.37169v1 Announce Type: cross Abstract: Mid-training equips pretrained large language models with specialized and reasoning capabilities, but the returns of this stage are bounded since add...