The $1/\mathcal{W}$ Law: Context Length is the Dominant Energy Lever in LLM Inference Fleets

Huamin Chen, Xunzhuo Liu, Yuhan Liu, Junchen Jiang and 3 more

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Beyond the Trial-and-Error Loop: Hybrid Projection and Automated Tuning for Distributed Training

Anshu Raina, Peyman Razaghi, Yuankai Chen, Cheng Yao and 6 more

Atlanta Poster Session 3, Thu, Dec 10, 10:00 AM–1:00 PM, Hall C1 · Published 2026

Learning from Disagreement: Maximum Divergence Knowledge Distillation

Aref Jafari, Parsa Ashrafi Fashi, Mehdi Rezagholizadeh, Hanieh Asadi Golmankhaneh and 4 more

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

SliMOO: Interpretable Multi-Objective Evolutionary Search for LLM Depth Pruning

Guanchen Li, Yixing Xu, Xuanwu Yin, Dong Li and 1 more

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

KroQuant: Kronecker-Structured Block Transforms for Efficient Post-Training Quantization of Diffusion Transformers

KroQuant applies a learned Kronecker-structured block transform to DiT activations for efficient W4A4 post-training quantization that outperforms SVDQuant and LoRaQ on image quality.

Yann Bouquet, Alireza Khodamoradi, Kristof Denolf, Mathieu Salzmann

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

DC-DiT: Adaptive Compute and Elastic Inference for Visual Generation via Dynamic Chunking

DC-DiT uses dynamic chunking to adaptively allocate tokens by region and timestep, reducing ImageNet inference FLOPs by up to 36.8% and improving FID by up to 37.8%. Its router enables elastic inference from a single checkpoint with smooth quality-compute tradeoffs.

Akash Haridas, Utkarsh Saxena, Parsa Ashrafi Fashi, Mehdi Rezagholizadeh and 2 more

Paris Poster Session 4, Thu, Dec 10, 5:30 PM–7:30 PM, Paris Poster Hall · Published 2026 · ▲ 16 on Hugging Face

LoRaQ: Optimized Low Rank Approximation for 4-bit Quantization

LoRaQ uses data-free optimization to quantize low-rank branches for 4-bit diffusion transformers, outperforming high-precision methods at equal overhead.

Yann Bouquet, Alireza Khodamoradi, Sophie Y Shen, Kristof Denolf and 1 more

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

TaskGround: Structured Executable Task Inference for Full-Scene Household Reasoning

TaskGround grounds full household scenes into task-relevant slices to infer executable task structures, improving compact open-weight models' success rates by large margins over direct prompting while cutting token costs up to 18x.

ZhiYuan Feng, Yu Deng, Ruichuan An, Zhenhua Liu and 10 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026