Frequently Asked Questions

Building methods that combine ontologies/knowledge graphs with LLMs/ML to produce verifiable answers, impact traces, and machine-usable schema updates. There are 3 tracks sharing the same ontology and seed data:

1. Track-1: NL → Ontology Querying

2. Track-2: Causal Reasoning (deterministic ± probabilistic)

3. Track-3: Ontology Evolution from integration data

No. It's a multi-week event: we open registration and answer questions at NeSy, and work continues online after the conference

Yes. Enter one, two, or all three tracks. Each track is judged independently.

1-5 members per team.

Students, researchers, and professionals worldwide (subject to applicable sanctions/export controls). You must be 18+.

Public, searchable support happens in Discord → #hackathon-tech-support. Private/sensitive issues → ask@optimas.ai

Register as an individual and join the Discord server.

Team creation/invites will be live soon. You'll see Create Team / Join Team on your dashboard when it goes live.

No—one team per participant (your team may submit to multiple tracks).

The team admin adds teammates by email. Invitees receive an email + dashboard notification; they must have an individual account to accept.

Track-1 (NL → Ontology Querying): Clear IR/plan/DSL with filters, joins, temporal windows, and alias/synonym handling. Answers should include evidence IDs (nodes/edges) + brief reasoning path

Track-2 (Causal Reasoning): Explicit propagation rules (directionality, cut-offs, time windows), readable impact trees (explain the “why”), and (optionally) probabilistic math with transparent confidence. Include one principled reverse-diagnosis example

Track-3 (Ontology Evolution): Clean field→class/property mapping, datatype/cardinality, anchors/placement (parent/extension), dedup/synonym merges, plus a machine-usable patch JSON (anchors, rationale + confidence, rollback/versioning notes).

Use the ontology tag for schema semantics (class/relation meanings, constraints, temporal scopes, aliases/synonyms), data for seed data formats/splits, and evaluations for metrics/baselines/reproducibility.

Use the provided seed data (and synthetic examples you create). Do not post or use private/sensitive data.

Yes. Disclose model names/versions, where inference runs (local vs provider), and any prompts/system instructions used.

No. Never share secrets or tokens publicly. For private debugging that requires sensitive info, email ask@optimas.ai.

A short write-up that covers reasoning rigor, AI/LLM technique & safety, feasibility & build plan, originality, and clarity. (See Judging section below for weights.)

1. Runnable demo/notebook + evaluation results
2. Repository Quality: LICENSE, README, setup steps
3. Video: 5-7 min + one-pager (method/results clarity)
4. Any items specified in the finalist brief

One Approach (Phase-1) per track. Finalists will receive build instructions and a submission portal for Phase-2.

Include your repo tag + commit SHA and exact script/command to reproduce.

Weighted auto-calc (reference):

0.30 x Ontology Reasoning Rigor
0.25 x AI/LLM Technique & Safety
0.25 x Feasibility & Build Plan
0.10 x Originality
0.10 x Clarity

Finalists are evaluated on Functionality & Rigor (runnable demo, results), Repository Quality, and Presentation (video + one-pager). Aggregate score + panel deliberation → Grand Prize per track.

1. Hard-coded IDs, one-off SQL per example, brittle rules that don't generalize
2. Opaque scores/confidence without explanation
3. Diagrams/architectures without executable logic or reproducible steps
4. Ignoring privacy / LLM disclosure requirements
5. Over-claiming without evidence/ablations/eval on provided datasets
6. Inventing ontology classes/relations or failing to ground in the provided schema

Unless stated otherwise, deadlines are AoE (Anywhere on Earth).

Key dates (summary):
Approach judging: Oct 22-24, 2025
Finalists announced: Oct 25, 2025 (webinar + site/Discord)
Build judging (finalists): Nov 25 - Dec 1, 2025
Winners announced: Dec 2, 2025

(Any changes will be posted in #announcements and on the Timeline page)

See the Judging page and announcements for track-specific awards and any partner prizes.

Participants must be 18+ and comply with local laws/sanctions/export controls. Organizers reserve the right to verify eligibility.

All submissions, code, designs, models, documentation, and other materials developed by participants during the Hackathon ('Submissions') are and shall remain the exclusive property of Optimas.ai Inc.

As a condition of entry, participants hereby assign all rights, title, and interest in Submissions to Optimas.ai, including all intellectual property rights worldwide.

Yes—see the Rules & Code of Conduct. At minimum:
Disclose LLM model(s), inference location, and prompts/system instructions.
No hidden human-in-the-loop steps for claimed automation.
No sensitive data, secrets, or PII in posts or repos.
Be respectful; zero tolerance for harassment.

In #hackathon-tech-support (Forum). Press New Post, pick a tag (track-1/2/3, ontology, data, evaluations), and follow the ✅ Post Guidelines.

Add the resolved tag (if enabled) or prefix the title with [Solved].

Yes—threads are public to participants so answers remain searchable. For private matters, email ask@optimas.ai.

Yes. Starters are in Python, but use what you prefer. Just ensure the repro instructions are clear and runnable

Yes, if it's executable and reproducible

1. Technical (public): post in #hackathon-tech-support with the right tag.
2. Private/sensitive or account/admin issues: ask@optimas.ai.
3. Logistics/timing updates: watch #announcements

We aim to reply within 1 business day (faster during sprint/judging windows).

Email ask@optimas.ai with a brief description—we'll route it to the right place.