DSOS

Sarawak State Legislative Assembly

Generate Optimal Seating Plan

IRL · LLM · MCTS · Feedback Loop — converged in under 5 minutes.

Total Seats

IRL Score

—

LLM Judge

—

MCTS Iter

Final

—

Run Optimization Pipeline

IRL learns history → MCTS searches → LLM judges → Best plan emerges

DUN Session

Iterations

IRL Discovered Weights

Hidden formula learned from 10 years of DUN session data.

System Settings

Configure the AI engine that powers DSOS

LLM Model Selection

Choose the language model that judges each candidate seating plan against DUN Standing Orders.

GPT-4o

OpenAI · best reasoning

Claude Sonnet 4

Anthropic · long context

Gemini 1.5 Pro

Google · multimodal

Llama 3.1 70B

Meta · open source

Mistral Large

Mistral AI · efficient

Qwen 2.5 72B

Alibaba · multilingual

Temperature

0.2 — lower = more deterministic

Max Tokens

LLM Prompt Engineering Manual

Upload the DUN Standing Orders PDF or your custom prompt manual. The system will parse and use it as the LLM judge's rulebook.

Drop PDF here or browse

PDF, TXT, MD, DOCX · max 25 MB

Or write/edit prompt directly

IRL · Past Years Training Data

Upload historical DUN session data (CSV/JSON/Excel). IRL will learn the hidden scoring formula from at least 10 years of records.

Drop files or browse

CSV, JSON, XLSX, ZIP · multiple files allowed

Years to use

Algorithm

Epochs

MCTS Engine Tuning

Engine parameters that rarely change once tuned. Iteration count is set per run on the Dashboard.

UCB constant (c)

Exploration vs exploitation balance (default √2)

Rollout depth

Sequential seat placements per simulation

Feedback Loop

After each real DUN session, retrain IRL with new speaking data and refine the LLM prompt automatically.

Enable automatic post-session feedback Auto-refine LLM prompt based on prediction errors

DUN Sarawak Chamber — Live Visualization

64 seats · semicircle layout · Speaker at front

No active arrangement

Click "Generate Optimal Seating Plan" to begin

Premier/Deputy Minister Govt Backbench Opposition Vacant

Generate Optimal Seating Plan

Run Optimization Pipeline

Plan Ready

Drafts (max 3)

IRL Discovered Weights

System Settings

LLM Model Selection

LLM Prompt Engineering Manual

IRL · Past Years Training Data

MCTS Engine Tuning

Feedback Loop

Saved Plans & History

DUN Sarawak Chamber — Live Visualization

Generate Optimal Seating Plan

Run Optimization Pipeline

Plan Ready

Drafts (max 3)

IRL Discovered Weights

System Settings

LLM Model Selection

LLM Prompt Engineering Manual

IRL · Past Years Training Data

MCTS Engine Tuning

Feedback Loop

Saved Plans & History

DUN Sarawak Chamber — Live Visualization

Draft Limit Reached