Claims Browser

A cost-efficient, hardware-accessible, and open-source pretraining recipe has been missing from the field

"a cost-efficient, hardware-accessible, and open-source pretraining recipe has long been missing"

FP8 precision training on consumer-grade RTX 5090 GPUs enables dramatic cost reduction for language model pretraining

"we train a collection of Puro-2B models from scratch on up to 1.4 trillion tokens with FP8 precision on consumer-grade RTX 5090 GPUs"

Language model pretraining has become almost synonymous with prohibitive cost, placing it out of reach for much of the academic and open-source communities

"Language model pretraining has become almost synonymous with prohibitive cost, placing it out of reach for much of the academic and open-source communities"

Patch-based processing is the practical unit for foundation model inference on whole-slide images due to their prohibitive size

"Whole-slide images (WSIs) are central to computational pathology but are prohibitively large, making patch-based processing the practical unit for foundation model inference."

I/O and orchestration overhead dominates end-to-end performance in large-scale WSI patch processing, not compute

"At scale, however, generating and handling massive numbers of patches on quickly introduces significant I/O and orchestration overhead, often dominating end-to-end performance."

Decoupling I/O, computation, and ingestion enables high-throughput WSI embedding extraction at scale

"We show that decoupling I/O, computation, and ingestion enables high-throughput WSI embedding extraction at scale."

WSI embedding extraction is fundamentally a data-centric systems problem rather than a compute-bound workload, with storage dominating beyond moderate concurrency

"By characterizing the scaling envelope, we demonstrate that storage dominates beyond moderate concurrency, reframing WSI embedding extraction as a data-centric systems problem rather than a purely compute-bound workload."

Distributed vector databases with metadata-coupled embeddings are particularly beneficial for low-resource environments in computational pathology

"This representation database is compact and reusable for tasks such as retrieval, classification, and few-shot learning, particularly benefiting low-resource environments."

Block drafters using all-parallel conditioning hit a fundamental information floor of 28.6% rejection at the final slot on Qwen3-4B, limiting maximum per-slot acceptance to 71%

"the all-parallel floor reaches $0.286$ at the final slot on Qwen3-4B, limiting even the best proposal to $71\%$ per-slot acceptance"

Adding just one realized token removes 86-100% of the information floor in block drafters, demonstrating strong locality in speculative decoding

"one realised token removes $86$--$100\%$ of this floor, a locality also recovered by an independent mutual-information analysis"

Current drafter models are far from optimal, with model gap accounting for 43-64% of DFlash rejection and 85-92% of DSpark's oracle-conditioned rejection at the final slot

"current drafters remain far above their floors: the final-slot model gap accounts for $43$--$64\%$ of DFlash rejection and $85$--$92\%$ of DSpark's oracle-conditioned rejection"

Block drafter rejection stems from two distinct sources that can be separated: missing within-block path information (information floor) and imperfect modeling of observable information (model gap)

"Their rejection mixes two losses: missing within-block path information and imperfect modelling of observable information"

About $4.4K is sufficient to reach the performance of Qwen2-1.5B based on the Puro Cost Scaling Law

"the fitted law suggests that about \$4.4K, less than \$5,090, is sufficient to reach the performance of Qwen2-1.5B"

Progress in vision generative AI has been driven by output quality, with hardware evolving reactively to accommodate growing model demands

"progress in vision generative AI has been driven by output quality, with hardware evolving reactively to accommodate growing model demands"

Vision generative models are equally needed in applications that operate under strict hardware constraints at the edge, including autonomous vehicles, agricultural sensors, and mobile devices

"vision generative models are equally needed in applications that operate under strict hardware constraints at the edge, including autonomous vehicles, agricultural sensors, and mobile devices"

A software-hardware co-design approach, where deployment constraints are considered from the start, will make generative AI deployment sustainable and accessible across a much broader range of platforms

"a software-hardware co-design approach, where deployment constraints are considered from the start of the design process, ensuring that the "right model" runs on the "right hardware" to serve the "right application", making generative AI deployment sustainable and accessible across a much broader range of platforms"

LLM compute accounts for 80-90% of query cost in production semantic data processing systems

"In production, LLM compute accounts for $80-90\%$ of query cost"

Each LLM call costs 100,000 to 10,000,000 times more than a relational predicate

"each call costs $10^5-10^7\times$ a relational predicate"