AI Video2026
Replaces a character in a video using a single repainted frame from the same clip and propagates that edit across the shot while preserving motion, camera and lighting; requires no pose estimator, segmentation, face tracker or text prompt. Key facts: a 33.1B MiniMax-H3 finetune, DMD-distilled to three forward passes, 124 frames in ~26s on one B200 GPU.
AI Video2026
Adds a plug-and-play linear-attention branch and LoRA adapters to MiniMax-H3 to run text-to-video generation faster than real-time (near-lossless quality tradeoffs). Includes an optimized FP8 inference stack and a community license with regional restrictions.
Uses lightweight discrete-diffusion adapters to draft blocks of tokens in parallel while keeping the original autoregressive model as the final arbiter, enabling lossless speedups (up to 3×) and compatibility with existing AR weights via a Diffusion Distillation phase.
Converts a natural-language function specification into a reusable local neural function by using teacher models to synthesize examples and finetuning a small adapter for a compact interpreter. Achieves higher semantic accuracy (83.6% on FuzzyBench-Hard) at the cost of roughly one minute compile time; produces versionable PAW artifacts for local deployment.
AI Video2026
Generates short multimodal videos from text, images, or reference clips using a fine-tuned MiniMax‑H3 fusion model; improves HDR clarity, motion fluidity, distant-face fidelity and VFX while preserving MiniMax‑H3’s prompt/style behavior. Best used via ComfyUI.
AI Video2026
LoRA adapters for MiniMax H3 that sharpen and enhance videos in ComfyUI by conditioning on source clips via guide latents for pixel-level alignment. Designed mainly for ref2va as a second-pass sharpening tool, includes a ComfyUI workflow and example before/after clips; requires aligned guide clips at the target resolution and valid clip lengths.
Investigates long-horizon memorization in language models by evaluating how complementary continual-learning mechanisms (data, function, weight anchors and low-rank allocation/LoRA) compose to reduce catastrophic forgetting across 100 sequential tasks; best composition raises final retention from 1.2% to 34.9%.
AI Model2026
Runs a 35B-class sparse MoE LLM with SSD-streamed experts, 4-bit quantization, prerouter routing prediction and Recover-LoRA adapters to enable ~2.9–3 GiB active memory and interactive decoding (~15 tok/s) for on-device inference.
AI Model2026
A reasoning‑efficient fine-tune of Qwen3.8-27B that penalizes overthinking tokens to shorten internal reasoning traces — about 58.3% fewer thinking tokens with <1% accuracy loss and ~1.95× speedup; designed for long-context, multimodal and quantized deployments.
Haojun Zhang, Yi Zou +8Affiliation: Xini Ding, Hao Li, Shuchang Zhou, Xianming Liu, Shiyu Huang‡, Affiliation: XPeng Inc.
Progressively prunes and distills audio encoders for speech LLMs to cut inference cost while preserving decoder-facing embeddings, using behavioral probes, representation alignment, cross-scale distillation and LoRA finetuning; reports reduced macro-error on Chinese–English benchmarks.
AI Audio2026
Encodes real audio into YuE2 semantic tokens and ships a matching NAR-branch LoRA so the YuE2 decoder renders realistic latents—enables tokenizing recordings, training artist LoRAs, and generating songs or covers. Includes an 8-layer tokenizer head and a rank-32 decoder LoRA; requires the YuE2 base models and GPU resources.
AI Model2026
Runs locally on constrained devices to turn text into guaranteed-parsable JSON tool calls, typed structured extractions, or sentence embeddings. Delivered as a single compact weights file (8–29 MB) with a laddered 2–20-layer design, low-bit quantisation and calibrated confidence scores for on-device apps.