Tech Times

LLM Data Mixture Breaks When Training Pools Shift: Causal Inference Offers Fix

LLM training data mixture optimization breaks when training pools shift — every prior proxy experiment becomes stale.
Tech Times

ICML 2026 Opens in Seoul Next Week: Record 23,918 Submissions Signal AI Agent Safety Era

ICML 2026 opens in Seoul on July 6 with a record 23,918 submissions — more than double last year — and a research program ...
IEEE

Federated Experiments: Generative Causal Inference Powered by LLM-Based Agents Simulation and RAG-Based Domain Docking

Abstract: Computational experiments method is an essential tool for analyzing, designing, managing, and integrating complex systems. However, a significant challenge arises in constructing agents with ...
IEEE

Characterizing Cloud-Native LLM Inference at Bytedance and Exposing Optimization Challenges and Opportunities for Future AI Accelerators

Abstract: As a major provider of LLM inference services, ByteDance has continuously explored diverse accelerator options to meet the rapidly growing inference demands of various heterogeneous LLM ...
GitHub

geo_inference_script.py

work_dir (str): The directory where the model and output files will be saved. mask_to_vec (bool): Whether to convert the output mask to vector format. mask_to_coco (bool): Whether to convert the ...
GitHub

deepgp_variational_inference_tutorial.py

from scratch in PyTorch. self-contained: no GPflow/GPyTorch. model: y = f_L(...f_2(f_1(x))) + eps, each f_l a sparse GP. inference: sample-through-the-layers variational inference.