Booth 21-25 | AI Data Management Zone | Tokyo Big Sight
Most improvements to AI systems happen before deployment: better training data, better fine-tuning, better RLHF. Once the model is out in the world, you generally get the performance you trained
Speed benchmarks in AI research can feel detached from anything practical. A paper claims a 10x speedup and you skim past it because the benchmarks never quite match the thing you’re actually trying to do. The FlexGNN result — up to 95.5x faster training on large graphs — is worth
