A pretrain-then-transfer method for streaming recommendation that decouples refreshable behavioral knowledge from task-specific geometry to enable continual model refresh without downstream interference; introduces Behavioral Multi-Token Prediction and Anchored Calibration Residual and shows 4–12% offline gains plus live Shopee A/B lifts.
A continuous-latent diffusion language model that preserves a high-capacity, decodable text latent and directly models its distribution via a block-causal diffusion transformer and query-based encoder–decoder; achieves top results on OpenWebText and XSum while scaling to 1B parameters.
Analyzes why supervised fine-tuning (SFT) causes severe task conflicts under multi-stage multi-task training while reinforcement learning (RL) enables stable coexistence, attributing the effect to sparse, near-orthogonal RL parameter updates and proposing Parallel-RL to decouple multi-task training.
AI Video2026
Experimental MiniMax H3 variant that injects learned stylistic and motion 'character' from LTX 2.3, Wan 2.2 and Krea 2 into H3 by surgically grafting attention and MLP components; preserves H3 modality routing while shifting t2v/i2v aesthetics, with limited audio impact and community-license constraints.
Systematically studies how language and vision interact during unified multimodal pretraining, identifies mechanisms that enable modality synergy versus competition, demonstrates the benefit of early joint training, and derives efficient pretraining recipes validated at scale.
AI Dataset2026
Provides ~39 TB of pre‑beamformed (channel capture) ultrasound RF data and metadata in zea/HDF5 format for reconstruction, flow, and inverse‑problem tasks. Released under CC‑BY‑4.0 and curated for training and evaluating ultrasound/RF foundation models.
AI Video2026
Packaged diffusers checkpoint of MiniMax H3 for image/text-to-short-video generation with native stereo audio; provided for direct use in diffusers image-to-video pipelines and aimed at easy integration into prototyping and production workflows.
AI Model2026
A MoE causal large language model for long-horizon agents, coding, and multi-step reasoning: 2.4T parameters (95B activated), native 262,144-token context (extensible to 1,010,000), multi-token prediction, and configurable thinking-mode reasoning controls.
Describes a 314B-parameter decoder-only Mixture-of-Experts language model that activates 13.2B parameters per token for fine-grained sparsity, long-context (up to 256K) and multi-domain capabilities. Emphasizes GDLA architecture, expert balancing, and multi-teacher distillation.
A research report proposing a continual-learning agent workflow that pairs recursive self-improvement with a Mixture-of-LoRA design: freeze a foundation model, compose specialist LoRA adapters routed per user turn, and support them with long-context RL and post-training infrastructure.
AI Agent Papers2026
Supports multimodal scientific understanding, long-horizon agentic workflows and scientific tool interaction using a unified pipeline of multimodal pretraining, supervised fine-tuning and scalable multi-task reinforcement learning. Distinctive features include time-series modules for signal forecasting and a separate Memory Decoder that enables rapid domain specialization without changing the frozen 397B backbone.
AI Agent Papers2026
Bobo Li, Hao Fei +31National University of Singapore 2University of Oxford, Project page: https://omni-scientist.github.io +1
Conducts end-to-end multidisciplinary research directly from heterogeneous raw evidence using lifecycle-wide perception and three autonomous agents (Ideation, Experiment, Writeup). Integrates perceptual analysis, execution provenance, and code-enforced checks to produce executable analyses, validated results, and compiled manuscripts across many modalities.