jill/LLMsForDisinformationPrediction

Files

T

History

WillJeynes f5f8800173 Update README, add final results, add util scripts to query based on user input.

2026-04-13 10:56:58 +01:00

..

.gitignore

Add a method of evaulation. Add back distilGPT version. Convert querying to another fastAPI

2026-04-10 19:23:41 +01:00

eval_interactive.py

Update README, add final results, add util scripts to query based on user input.

2026-04-13 10:56:58 +01:00

eval.py

Add a method of evaulation. Add back distilGPT version. Convert querying to another fastAPI

2026-04-10 19:23:41 +01:00

full.py

Add deepseek version, full trained version. Add results

2026-04-11 12:02:18 +01:00

lora2-1.py

add alternate training

2026-04-13 09:57:48 +01:00

lora2.py

Use mini llama, getting more interesting results

2026-04-10 17:34:36 +01:00

lora23.py

Add deepseek version, full trained version. Add results

2026-04-11 12:02:18 +01:00

lora.py

Add a method of evaulation. Add back distilGPT version. Convert querying to another fastAPI

2026-04-10 19:23:41 +01:00

q_full.py

Add deepseek version, full trained version. Add results

2026-04-11 12:02:18 +01:00

q_lora2.py

Update README, add final results, add util scripts to query based on user input.

2026-04-13 10:56:58 +01:00

q_lora23.py

Add deepseek version, full trained version. Add results

2026-04-11 12:02:18 +01:00

q_lora.py

Add a method of evaulation. Add back distilGPT version. Convert querying to another fastAPI

2026-04-10 19:23:41 +01:00

README.md

Update README, add final results, add util scripts to query based on user input.

2026-04-13 10:56:58 +01:00

requirements.txt

Add a method of evaulation. Add back distilGPT version. Convert querying to another fastAPI

2026-04-10 19:23:41 +01:00

README.md

Results

Model/Technique	Coherence	Plausibility	Disinformation?
distilGPT2 + LoRa	6/9	4/9	2/9
miniLLama + LoRa	7/9	6/9	5/9
deepseek + LoRa	7/9	5/9	5/9
distilGPT2 (full training)	4/9	3/9	2/9
miniLLama + LoRa (rotate q + multigen)	8/9	8/9	7/9