Explore tools, prompts, skills, news and resources related to ACT 第28页.
Prompt works related to ACT 第28页
Latest articles, prompts, use cases and editorial content about ACT 第28页
FOSCAR is a GPU kernel library that speeds up decode attention over 2-bit and 4-bit quantized KV caches by applying scale and offset through affine folding rather than per-value conversion. It uses bit manipulation and w...
MemoryDojo is a benchmark for tracing how poisoned information propagates through LLM agent memory systems across sessions and measuring its influence on later agent behavior. It comprises 155 multi-session long-horizon...
The paper proposes Intent-Aware Caching (IAC), an online eviction policy for key-value prefix caches in LLM serving systems. It classifies incoming requests by user intent, learns lightweight reuse statistics from recent...
MIT Technology Review argues that opposition to AI data centers is driven less by a vague culture war than by the physical consequences of enormous electricity demand. Reporting from Virginia's Data Center Alley des...
Researchers propose ICare-DT, a Transformer-based digital twin that updates a latent patient state hourly from new intensive care measurements without retraining. Trained on 140,473 eICU-CRD stays, later forecasts improv...
The paper proposes Head-Adaptive Sparse Attention (HASA), a training-free sparse attention framework for LLMs that adapts selection resolution along feature and sequence dimensions. It uses each pretrained head's qu...
Baidu announced ERNIE Evolve, an automated research system that proposes research directions, edits AI training and inference code, runs experiments, and keeps changes only after passing a hidden evaluator and reviewer....
Google announced TPU 8i Max and TPU 8tx as new system configurations within its eighth-generation TPU lineup. TPU 8i Max packages two TPU 8i chips to double compute, SRAM and HBM bandwidth in the same rack footprint for...
SURE is a data-efficient training framework for vision-language-action policies that explicitly models spatial and temporal uncertainty in manipulation demonstrations. It adds a lightweight transposed-convolutional proba...
This paper investigates how reward models learned from human preferences can be exploited by sampling-based control methods like MPPI and best-of-N selection. Experiments on HalfCheetah, Hopper, and Meta-World tasks show...
Researchers propose RelGuard, the first differentially private framework for GraphRAG, offering formal privacy guarantees against knowledge graph reconstruction attacks. It introduces relation differential privacy (relat...
This paper studies private-data leakage in federated learning when Vision Transformers are fine-tuned with LoRA, where low-rank updates weaken the gradient signal available to the server. It proposes a spatially anchored...
The paper introduces Test-Time Model-Based Reinforcement Learning (TT-MBRL), which uses gradient descent at test time to encode episode observations into world-model parameters. This gives the agent a distinct world mode...
AbCD converts multi-agent debate traces into preference data without gold-answer labels. It clusters final answers by task-level semantic equivalence and combines answer equivalence with confidence-weighted disagreement....
Researchers propose QGPM, a training-free method that reshapes rotary position embeddings using query relevance derived from the model's own attention heads. It builds a broad relevance map for lower layers and a sh...
Atik is an open-source RISC-V RocketChip accelerator that executes transformer matrix multiplication and attention on a shared fixed-point MAC mesh. It uses streaming online-softmax to avoid materializing the full score...
V-STORM is a framework that converts videos into Spatio-Temporal Scene Graphs and derives Grounded Chain of Thought annotations, linking each reasoning step to explicit spatio-temporal evidence. It includes a 260K-sample...
Researchers introduce e-means, a mini-batch k-means procedure using an exponential moving average optimization scheme and scaling laws for decay rate and batch size. The paper reports that e-means reaches Lloyd k-means&#...
LEMMO is a large electromagnetic model for instruction-driven recognition and description of long raw IQ recordings. It combines a linear-attention EM encoder with anti-aliased temporal resampling to process up to eight...