Zenn

2026-05-21

Measuring the "emotional intelligence" of a local LLM — Benchmarking Qwen3.5-122B(int4) with EQ-Bench 3

To begin with, when evaluating LLM, knowledge and inference benchmarks such as MMLU and GSM8K are standard, but we also consider questions like, "How well can it interpret interpersonal nuances?" and "...

Zenn

2026-04-12

Creating an LLM that answers using only hiragana with Groq and DPO - From synthetic data generation to training and evaluation -

Initially, even when instructing LLM to "answer using only hiragana," it's common to receive answers that include kanji and katakana. This is because children's education...

Zenn

2026-04-08

Deploying Qwen3-Omni with two RunPod H100s.

First, Qwen3-Omni-30B-A3B-Instruct is a native omni modal model developed by Alibaba Cloud…

Zenn

2026-04-02

Setting up NemoClaw on a DGX Spark via remote access

First, we'll remotely access NVIDIA's publicly available AI agent framework, NemoClaw, using Tailscale...

Zenn

2026-02-26

[MLflow] A Guide to Building a Secure LLM Learning Management Environment with Tailscale + AWS EC2/S3

First, when fine-tuning an LLM (Large-Scale Language Model), it is necessary to consider the hyperparameters and loss progression for each experiment, as well as model saving (Artifact...).

Zenn

2026-02-26

[MLflow] Setup Guide: Building a Secure LLM Training Management Environment with Tailscale + AWS EC2/S3

When fine-tuning LLMs (Large Language Models), managing hyperparameters, loss trends for each experiment, and model saving (Artifa...

Zenn

2026-02-04

Running torchaudio with DGX Spark (sm_121)

Hello! My name is Fukazawa, and I've been interning at Quixotiks Inc. since October 2024. Currently, I'm working on NVIDIA DGX Spa…

Zenn

2026-02-04

Running torchaudio on DGX Spark (sm_121)

I'm Fukazawa, an intern at Quixotiks since October 2024. Currently, I'm working on NVIDIA DGX Spa…

Tech

2026-01-06

How to run torchaudio with Docker+GPU on DGX Spark

Hello! I'm Fukazawa, an intern at Quixotiks Inc. since October 2024. I am currently working on speech recognition using NVIDIA DGX Spark...

Tech

2025-10-29

Facial expression dataset specialized for Japanese and East Asians

I researched and compiled information on datasets specifically tailored for Japanese (Asian) individuals and the elderly.

Tips

2025-10-14

How does AI interpret "Grandpa's grumpy face"? Testing VLM's facial expression recognition with elderly Asians.

We decided to investigate how accurately the latest VLMs can recognize facial expressions in Asians (Japanese) and the elderly, who can be considered "outliers" in typical datasets.

2025-10-03

How do VLMs describe human facial expressions? A comparative experiment with Gemini and GPT-4o.

By having the model describe human expressions, we aimed to clarify whether it "cannot recognize faces at all" or "can recognize faces but cannot classify the subtle nuances of expressions."

LLM

2025-09-25

Facial expression classification with the "In the Wild" dataset

I will report on the experimental results of how well the latest Vision-Language Models (VLMs), Gemini and GPT, can recognize facial expressions, using the In the Wild dataset we previously investigated.

tips

2025-09-22

"In the Wild": Video Datasets for Facial Expression Recognition

Once again, I'm reporting on a technical survey concerning Facial Expression Recognition (FER).

2025-08-16

Key Classifications of Facial Expression Recognition Datasets: Controlled and In the Wild

I've been involved in research work, and I'm now looking forward to sharing our survey findings in articles.