BLOOM: A 176B-Parameter Open-Access Multilingual Language Model
BLOOM is a 176B-parameter open-access language model designed and built as a step towards democratizing powerful language technology. It achieves competitive performance on a wide variety of benchmarks, with stronger results after undergoing multitask prompted finetuning.
BLOOM is a powerful open-access language model that businesses can take advantage of to improve their natural language processing capabilities.
Large Language Models with Controllable Working Memory
The paper proposes a novel method - Knowledge Aware FineTuning (KAFT) - to strengthen both controllability and robustness by incorporating counterfactual and irrelevant contexts to standard supervised datasets. The evaluation shows the utility of KAFT across model architectures and sizes.
KAFT can improve the controllability and robustness of LLMs, making them more reliable for businesses to use in their natural language processing workflows.
Efficiently Scaling Transformer Inference
The paper studies the problem of efficient generative inference for Transformer models and develops a simple analytical model for inference efficiency to select the best multi-dimensional partitioning techniques. The approach achieves a new Pareto frontier on the latency and model FLOPS utilization tradeoffs on 500B+ parameter models that outperforms the FasterTransformer suite of benchmarks.
Efficiencies in generative inference for Transformer models can improve their scalability and performance, making businesses more efficient in their natural language processing workflows.