ThaiLLM-8B-MedApp

ThaiLLM-8B-ToolUse is a reinforcement learning fine-tuned version of ThaiLLM/ThaiLLM-8B-IQ, trained specifically for routing a user's request to the correct medical tool call and information query.

Training Details

The model was trained using Prime-Intellect's prime-rl framework with the same configuration as ThaiLLM/ThaiLLM-8B-ToolUse

Results

Model	Response (LLM Judge)	Citation (F1)	Accuracy	Trigger (F1)	Macro (F1)
typhoon-s-thaillm-8b-instruct-research-preview	60.80	11.70	67.50	47.50	39.37
Qwen3-30B-A3B	—	—	99.00	99.20	97.76
ThaiLLM-8B-IQ	69.30	70.00	—	—	—
ThaiLLM-8B-ToolUse	—	—	99.90	100.00	99.28
ThaiLLM-8B-MedApp	68.33	68.40	98.63	100.00	89.64

Per Tool F1 Results

Tool	typhoon-s-thaillm-8b-instruct-research-preview	Qwen3-30B-A3B-Thinking-2507	ThaiLLM-8B-ToolUse	ThaiLLM-8B-MedApp
create_appointment	7.07	98.67	99.50	98.83
create_reminder	35.97	98.84	100.00	51.12
get_health_emergency_contact	30.26	99.39	100.00	99.54
list_appointment	51.87	99.35	98.06	97.88
list_reminder	56.35	94.04	98.68	70.64
no_tool	76.59	99.33	100.00	100.00
prescreen	5.13	93.44	98.08	99.19
search_medical_facts	51.71	98.98	99.92	99.91