Observed Signal · Sep 30, 2026 · Corporate Update · Source: Appier · Impact: 4/5

Appier Research Accepted at NeurIPS: AI Agents Learn Not Only to Use Tools, but to Build Their Own

Executive Signal Summary

Appier (TSE: 4180) announced that its research paper on the SMITH framework has been accepted at NeurIPS. The framework lets AI agents create and use tools in a single reinforcement-learning loop, and can be shared across models of different sizes. According to the release, a roughly 4-billion-parameter model trained with SMITH outperformed a baseline using a ~30-billion-parameter model, and token use for repeated reasoning fell from 3,206 tokens to about 100 tokens — a ~32-fold efficiency gain.

Polaris7 AgentPolaris7 Strategic Assessment
High Confidence

The NeurIPS acceptance of Appier's SMITH framework strengthens its position as a research-driven Agentic AI company. Demonstrating that small models can build reusable tools with high token efficiency is strategically relevant for scalable multi-agent collaboration in AdTech and MarTech enterprise workflows.

MARKET SIGNALS

Track Appier Signals & Strategic Shifts

Polaris7 connects regulatory filings, primary sources, leadership changes, and M&A activity into structured intelligence. Set up your Polaris7 workspace to monitor relevant companies in your executive briefing.

  • Verified primary sources
  • Automated briefings
  • Instant watchlist setup

Key Takeaways & Evidence Grounding

  • Appier (TSE: 4180) announced on October 1, 2026 that its paper 'Joint Optimization of Tool Creation and Use for Large Language Model Agents' was accepted at NeurIPS, the premier AI and machine learning conference.
  • The paper introduces SMITH (Schema-grounded Multi-task Iterative Tool Honing), a reinforcement learning framework that combines tool creation and tool use in one training loop.
  • A model of about 4 billion parameters trained with SMITH built tools that outperformed every other method in the study on unseen tasks, including a baseline where a roughly 30-billion-parameter model built tools on the fly.
  • SMITH reduced average output from 3,206 tokens with conventional step-by-step reasoning to about 100 tokens, a roughly 32-fold efficiency gain.
  • Tools built by the small model worked across model sizes, including a lightweight model of about 350 million parameters, and could be shared across models and tasks.
  • The research was highlighted by Appier CEO and Co-founder Dr. Chih-Han Yu and Research Scientist Chieh-Yen Lin.

Connected Companies & Entities

“Appier (TSE: 4180) is an AI-native company delivering Agentic AI as a Service (AaaS); it announced the NeurIPS-accepted SMITH research paper...”

Primary Source Grounding & Direct Attribution
Direct Origin Attribution
Primary Reporting: Appier•Published: Sep 30, 2026

Related Market Signals & Shifts

Recent verified developments and strategic activity across this market segment.

CybersecurityDec 22, 2025

Import AI: Cyber AI Overhang and New Research Tools

This Import AI newsletter issue argues AI progress is increasingly powerful yet often invisible to most people, creating a growing “cyber-AI capability overhang.” It highlights new research showing that when large language models are placed inside scaffolding frameworks they reveal stronger cybersecurity abilities: ARTEMIS, a multi-agent penetration-testing scaffold developed by researchers (Stanford, Carnegie Mellon, Gray Swan AI), significantly outperformed other agent scaffolds in a realistic university-network red-team exercise and matched or exceeded typical professional performance at lower API cost. The issue also summarizes OSMO, an open-source tactile glove co-developed with Meta researchers that improves human-to-robot skill transfer, and ChipMain/ChipMind, tooling that converts chip specifications into a knowledge graph (ChipKG) to let LLMs reason about complex semiconductor designs, achieving strong benchmark results on SpecEval-QA. The piece frames these findings as evidence that modern AI is under-elicited and that elicitation frameworks, tooling and infrastructure matter for real-world impact.

Read assessment
Large Language Models (LLM) & AIMay 24, 2026

AI Mathematics Breakthrough and Multi‑Agent Lab Result

Exponential View reports two recent AI research milestones: an OpenAI reasoning model produced a proof that resolves an 80‑year‑old open problem in discrete geometry, a result validated by human mathematicians who described the discovered connection as unexpected; and a multi‑agent system named Robin completed an end‑to‑end scientific cycle (hypothesis → experiment selection → data analysis → refinement), enabling human-run wet‑lab experiments that identified an existing drug candidate for repurposing to treat macular degeneration. The newsletter argues these developments illustrate AI’s potential to bridge isolated scientific domains and to accelerate the full hypothesis‑experiment‑analysis loop in research.

Read assessment
AI AgentsOct 4, 2026

Personal AI Agents Map Emerges as New Delegation Layer

The article maps the emerging personal AI agents landscape as of late 2026, highlighting a shift from chat-based assistants to autonomous agents that act on behalf of users. Key developments include major platform launches: Meta's Muse, Apple's Siri AI beta, Anthropic's Cowork in Claude, Rabbit OS3, Microsoft Autopilot, OpenAI Dots, Google Gemini Spark, and Amazon's shopping agent in Alexa. Instinct, a startup, raised $1B at a $10B valuation. The author argues these agents constitute a new 'delegation layer' between user intent and fulfilment, potentially more valuable than search or checkout. The article is largely paywalled, with the core analysis behind a subscription.

Read assessment

WATCHLIST & BRIEFINGS

Receive future signals for Appier & competitors in your briefing

Stay ahead of material M&A deals, leadership changes, account wins, and product rollouts. Polaris7 connects primary filings, company newsrooms, and trade reporting directly into your executive briefing.

  • Regular primary-source checks
  • Structured Watchlist briefings
  • Preserves current context