Simple Attention
AssessTechniques
An attention block architecture used in Whistle's encoder and decoder.
Why it's here
Placed in Assess: 3 article(s) of evidence from 1 source(s), led by model releases, with 1 in the last 30 days. Confidence 39%.
Evidence (3)
- 7Hacker News·10/8/2026model_releaseWhistle: 16.9 MB On-Device Speech-to-Text Model
Whistle is a compact speech recognition model designed to run entirely on-device on CPUs with no dependencies. It supports transcription, word-level timestamps, and speech embeddings for up to 30 seconds of audio in seven languages, with audio never leaving the device.
- 3Hacker News·8/20/2026researchWhy Turns Can Beat Radians
The article argues that using turns, a [0,1] circle-based scale, can simplify trigonometric code by avoiding repeated conversions to and from radians. It also notes that common fractional turn values are often represented more exactly in floating-point than radian equivalents, which can improve precision and reduce unnecessary computation.
- 7Hacker News·8/10/2026model_releaseNeedle 2: a 14MB agentic LLM for edge devices
Cactus released Needle 2, a compact 14MB agentic LLM designed for tool use, structured extraction, and device control on phones, wearables, smart home devices, microcontrollers, and small robots. The model uses 45M parameters at 2-bit compression, claims very high token throughput on edge hardware, and can be fine-tuned locally with a Python package and automated data generation pipeline.