The study evaluates the portability of INT8 post‑training quantization across seven hardware platforms, including CPUs, GPUs, and vendor NPUs, by keeping the ONNX model and quantization scales constant. It finds that INT8 performance and output consistency vary significantly: CPU dot‑product instructions determine speedup, identical INT8 outputs only occur when integer kernels match, and vendor NPUs require their own quantization pipelines. The authors also show that edge‑NPU latency is dominated by data transfer rather than compute and provide scripts and reports for reproducibility.
By Yuyeong Shin
arXiv:2608. 13756v1 Announce Type: new Abstract: Two GPU kernels implementing the same scaled INT8 GEMM interface are usually treated as interchangeable.
By Teng-Ruei Chen
arXiv:2609.39223v2 Announce Type: new
Abstract: Large language model (LLM) inference is increasingly moving toward lower precision to realize the throughput of hardware accelerators, but aggressive p...
By Weili Xu, Jisen Li, Yuqing Jian, Chenxi Li, Zhizhou Sha, Yifan Yu, Qingyang Wu, Chenfeng Xu, Zhongzhu Zhou, Tianyi Zhang, Ben Athiwaratkun
arXiv:2606. 19535v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly deployed in sensitive settings such as software engineering, where their outputs directly shape downstream artifacts.
By Nils Loose, Jonas Sander, Felix M\"achtle, Thomas Eisenbarth
The paper introduces the Golden Ruler, a catalog of 109 numeric formats for machine learning hardware, including FP8, BF16, MXFP4, and microscaling block formats. It provides six bit‑exact conformance packs that map to IEEE P3109 v3.2.0 standards, each as a self‑contained JSON document with a SHA‑256 fingerprint and an anchor vector for cross‑pack sanity checks. The work cross‑validates these packs against ml_dtypes 0.5.4 and documents any divergences as spec‑permitted gaps, offering a vendor‑neutral reference for engineers.
By Dmitrii Vasilev
arXiv:2606. 09686v1 Announce Type: cross Abstract: Numeric format proliferation in machine learning hardware -- FP8 (E4M3 and E5M2), BF16, MXFP4, microscaling block formats, and dozens of research variants -- has outpaced the availability of vendor-neutral, bit-exact reference material.
By Dmitrii Vasilev
The paper evaluates the effectiveness of tolerance‑based conformance tests for INT8 quantized GEMM kernels used in large language models. By injecting nine faults into a Qwen3‑1.7B reference pipeline, the authors show that most faults shift outputs by at most one bfloat16 spacing, rendering a tolerance of one spacing blind to these errors. They further demonstrate that requantizing weight scales to the nearest power of two aligns CUTLASS and Triton implementations bit‑for‑bit and produces identical token sequences, with only minor perplexity changes.
By Teng-Ruei Chen
arXiv:2607. 14431v1 Announce Type: cross Abstract: We report a way to make a frozen small language model both more capable and dramatically cheaper at once, without changing any weights.
By Sietse Schelpe
arXiv:2607. 14568v1 Announce Type: cross Abstract: A companion study ran a 35B mixture-of-experts model on a 2011 NVIDIA Tesla C2075 (Fermi, sm_20, 6GB) as a GPU-prefill/CPU-decode hybrid, because the 4-bit model did not fit in device memory (arXiv:2606.
By A. C. Opus, J. Q. Lu
arXiv:2608. 08910v1 Announce Type: cross Abstract: PTQTP decomposes LLM weight matrices into two ternary (trit) planes with two free per-group scales.
By Matteo Grella
The paper investigates how exporting ternary language models (BitNet, Falcon‑E, BitCPM) through a bf16 cast step can introduce significant discrepancies between the fine‑tuned latent weights and the deployed ternary codes. In three lab pipelines, the authors find that fp32 quantization of shipped latents disagrees with the deployed codes on up to 1.77% of codes, and that the export step can drastically reduce strict accuracy on GSM8K (e.g., from 58.79% to 0.78% for Falcon‑E‑1B‑Base). They propose two compatibility remedies—directly writing the training quantizer’s codes or adjusting bf16 inputs—to meet a 4‑point strict‑accuracy non‑inferiority criterion across all models.
By Avichal Sahai (Ofbusiness), Nishant Raj (Ofbusiness), Animesh Srivastava (Ofbusiness)
arXiv:2606. 00279v1 Announce Type: cross Abstract: Verifying claims about AI workloads is a pre- requisite for credible AI governance of covert adversaries (who comply with monitoring only when detection likelihood is high), yet the ap- parent non-determinism of GPU floating-point arithmetic forces auditors to accept approximate output matches.
By Naci Cankaya