The MiCA (Minor Component Adaptation) method has been merged into the main branch of the Hugging Face PEFT library, allowing users to install it directly from source. It is exposed through the existing LoRA interface by setting `init_lora_weights="mica"`.
A Reddit user compares the pricing and specifications of the AMD Instinct MI210 64GB and the Chinese DCU K100 64GB GPUs available on the Chinese eBay market. The discussion highlights that while both cards offer similar memory capacities, they differ significantly in price, bandwidth, and architectural details.
A manual experiment tested whether a procedural scaffold generated by a large model can transfer planning discipline to smaller models without fine-tuning or revealing the target answer. The results indicate that this approach significantly improves structural readability and composition in small models when applied across different Three.js domains.
This study investigates the statistical learning and mental representation of neural language models by training Generative Transformer models on a synthetic grammar and analyzing their internal representations at various stages.
This article identifies a distinct failure mode in large language model agents where they struggle to discard outdated facts in favor of current ones, even when comprehension is intact. The authors demonstrate that this "supersession gap" persists across model scales and memory sizes, indicating it is a trainable bottleneck rather than a limitation of context window or model strength.
The llama.cpp project has released version b9838, providing pre-built binaries for a wide range of operating systems and hardware accelerators. This release includes support for CPU, GPU (CUDA, Vulkan, ROCm, OpenCL), and specialized AI accelerators across macOS, Linux, Windows, Android, and openEuler.
This work introduces Aloe-Vision, a family of open-source large vision-language models (7B and 72B) trained on the newly released Aloe-Vision-Data dataset to address data scarcity and robustness issues in healthcare AI. The authors demonstrate that their high-quality training mixture yields significant performance gains over baselines while maintaining general capabilities.
A re-derivation of the activation patching estimand from causal mediation analysis reveals that the natural indirect effect (NIE) captures not only the causal effect through a specific component but also interaction effects (INT). These INT terms measure how much a component's causal effect depends on the state of other components in the model, challenging the assumption that NIE isolates individual contributions.
The authors introduce the context-ready transformer, a recurrent neural network architecture that pre-contextualizes each token before it enters a D-layer transformer block using a correction network.
The authors propose Entropy-guided Multi-Token Prediction (EntMTP), a training-free scheduler that dynamically adjusts speculation depth during LLM inference based on local generation entropy. This approach addresses the inefficiency of static tree-based attention topologies by matching compute requirements to context predictability.
The article introduces Ko-WideSearch, a new benchmark designed to evaluate the breadth-search capabilities of web agents in Korean, addressing the lack of exhaustive set enumeration metrics outside English.
The authors introduce Narrative-UFET, a controlled extension of ultra-fine entity typing that pairs entity mentions with automatically generated short narratives to address limitations in long-tail type disambiguation. The study demonstrates that narrative context yields consistent improvements over sentence-level baselines, particularly when the entity's type shifts within the text.
The authors introduce Masked Language Flow Models (MLFMs), which combine masked diffusion with continuous flows to enable efficient, multi-step reasoning in language generation. This approach bridges the gap between parallel generation efficiency and complex task performance by allowing pretrained models to be adapted into MLFMs.
This paper introduces DysLexLens, a low-resource LLM framework designed to analyze the experiences of dyslexic learners with AI tools through online forum discussions. The system provides an end-to-end, evidence-traceable architecture that transforms noisy social media posts into focused corpora and generates verifiable query responses.
This paper addresses the performance degradation of offensive comment detection models when deployed across different Chinese social media platforms by proposing a dual-threshold hard example mining method.
The Generational Context Architecture (GCA) proposes treating an LLM's context window as a finite lifespan rather than infinite storage to solve "context rot" and attention dilution in multi-agent systems. By enforcing artificial mortality, agents are terminated before performance degrades, passing their state to new generations via a flat-file Markdown vault.
The Yuvion LLM is a new large language model designed to address safety failures by treating adversarial robustness and agentic capability as primary objectives. It utilizes a pipeline combining adversarially aware data construction, knowledge-enhanced continued pretraining, and policy-grounded multi-task safety post-training.
The authors introduce DiscoBench, a benchmark designed to evaluate whether search agents powered by large language models can proactively identify ambiguity and ask effective clarification questions during deep search tasks. Unlike existing benchmarks that assume complete user queries, this framework addresses the reality of vague or underspecified requests in real-world scenarios.
This study conducts a factorised analysis of probe-based uncertainty estimation to determine what drives performance in detecting hallucinations within Large Language Models. The research isolates variables across feature design, training data, and evaluation settings to provide clear insights into effective methodologies.
This article addresses the issue of unidentifiable latent states in LLM-based world models caused by history bypass, proposing strict latent state mediation to resolve this. The authors introduce textual latent states and factorized GRPO (fGRPO), a tree-structured reinforcement learning method that enforces strict mediation during training.