dataXRESEARCH
DATAX AI SCIENCE × TYPHOON R&D

Intelligence,
grounded in
research.

Advancing language models and the agents built around them. From Thai language understanding to memory, reasoning, and reliable real-world interaction.

Built on collaboration. Grounded in evidence.
LANGUAGE MEMORY AGENTS EVALUATION
CONNECTED IDEAS.
NEW POSSIBILITIES.
[ DX / 01 ]
A shared research
perspective.
29selected publications
12featured researchers
06connected research areas
RESEARCH COLLECTION2025 — 2026

01 / RESEARCH DIRECTIONS

From foundations
to capable systems.

We connect model research with the questions that matter in practice: what models understand, how agents act, and how we know they work.

02 / CURRENT EXPLORATIONS

The questions
we’re working on next.

RESEARCH IN PROGRESS

New directions in how models study, how agents behave, and how their capabilities are evaluated.

01

Machine
studying

Our emerging research direction on machine studying, alongside broader work on model knowledge and memory.

KNOWLEDGE × MEMORY
02

Automatic
agentic evaluation

Investigating automated ways to assess agent behavior, tool use, and task outcomes across realistic interactions.

MEASUREMENT × RELIABILITY
03

Agent
simulation

Exploring simulated interactions as a way to study agent behavior and human–agent workflows.

INTERACTION × BEHAVIOR
04

Role-play
LLMs

Studying role consistency, dialogue behavior, and tool-use decisions in role-conditioned language-model agents.

ROLES × TOOL USE

Active directions shared by the research team. These are ongoing investigations, not claims of completed results or released systems.

03 / SELECTED PUBLICATIONS

Ideas, tested
and shared.

Explore selected co-authored work across DataX, Typhoon, and the wider research community. Every paper links to its source.

Featured first · then newest

Includes collaborative work and related research by featured authors; not all papers were produced under the same institutional affiliation. Venue badges distinguish proceedings, workshops, and preprints. SEATauBench’s acceptance is reported on the author’s website.

04 / PEOPLE & COLLABORATION

Different perspectives.
Shared curiosity.

A connected group of scientists and collaborators bringing together language, learning, and systems research.

DataX AI Science & contributors

MODELS · KNOWLEDGE · AGENTS

Typhoon R&D & contributors

LANGUAGE · REASONING · MULTIMODAL

A selected research network, not a complete staff directory. Core groups follow the supplied research overview; additional contributors are included through public team and publication records.

05 / OPEN RESEARCH

Take a closer look.

Read the work. Explore the resources. Build on the ideas.

GOOD RESEARCH STARTS WITH A CONVERSATION.

Let’s ask the
next question. Together.

Interested in language models, agent evaluation, memory, or Thai AI? Connect with Saksorn to start a research conversation.