Text-to-Video Diffusion Model Resilience Study
On the Resilience of Text-to-Video Diffusion Models to Hardware Faults
The study reveals reliability risks in deployed T2V systems, particularly with bfloat16 format, affecting video generation quality and semantic accuracy.
Researchers conducted the first systematic study on text-to-video diffusion model resilience to hardware faults. The fault-injection study covered three T2V models and a benchmark, showing a single fault can degrade performance by up to 3.7%. Memory faults proved more damaging than computational faults, with 7-28% of faults causing visible artifacts including semantic changes.
On the Resilience of Text-to-Video Diffusion Models to Hardware Faults
We present the first systematic study of the resilience of text-to-video (T2V) diffusion models under random hardware-level faults. While T2V models are widely used for automated video generation due to their ability to produce high-quality, temporally coherent, and realistic videos, their iterative denoising process and spatiotemporal dependencies introduce unique failure modes. We perform an extensive fault-injection study covering both computational and memory faults across three T2V models and a representative benchmark. Our results show that (1) a single fault can degrade overall performance by up to 3.7\%, with semantic correctness more affected than perceptual quality; (2) memory faults are more damaging than computational faults, high-order exponent bits are particularly vulnerable, and the widely-used bfloat16 is more susceptible than alternative formats; and (3) 7-28\% of faults cause visible artifacts, including semantic changes such as added objects, suggesting that single faults are sufficient to alter output semantics. Our findings reveal reliability risks in deployed T2V systems and motivate further research on improving fault resilience. Code: \href{https://github.com/ztcoalson/T2V-Resilience}{https://github.com/ztcoalson/T2V-Resilience}.
