A collection of models and dataset from the paper "The Hallucination Tax of Reinforcement Finetuning".
AI & ML interests
Natural Language Processing
Recent Activity
Organization Card
LIME NLP is part of the USC NLP Group. Our team's primary focus is on creating trustworthy NLP models. We meticulously investigate the ethical consequences and broader societal effects of NLP models, striving to ensure that language technologies are constructed and employed in ways that align with ethical guidelines and uphold human values.
models
20
lime-nlp/Llama-3.1-8B-Instruct-SUM50
8B
•
Updated
lime-nlp/Llama-3.1-8B-Instruct-SUM30
8B
•
Updated
lime-nlp/Llama-3.1-8B-Instruct-SUM10
8B
•
Updated
•
2
lime-nlp/Llama-3.1-8B-Instruct-SUM01
8B
•
Updated
•
4
lime-nlp/Llama-3.1-8B-Instruct-SUM00
8B
•
Updated
lime-nlp/Qwen2.5-7B-SUM00
8B
•
Updated
•
3
lime-nlp/Qwen2.5-7B-SUM01
8B
•
Updated
lime-nlp/Qwen2.5-7B-SUM10
8B
•
Updated
lime-nlp/Qwen2.5-7B-SUM30
8B
•
Updated
•
1
lime-nlp/Qwen2.5-7B-SUM50
8B
•
Updated
datasets
7
lime-nlp/OSReward
Preview
•
Updated
•
3
lime-nlp/Synthetic_Unanswerable_Math
Viewer
•
Updated
•
36.8k
•
53
•
14
lime-nlp/DeepScaleR_Difficulty
Viewer
•
Updated
•
5.06M
•
80
•
9
lime-nlp/orz_math_difficulty
Viewer
•
Updated
•
6.18M
•
30
lime-nlp/MATH_Difficulty
Viewer
•
Updated
•
1.61M
•
47
lime-nlp/GSM8K_Difficulty
Viewer
•
Updated
•
1.13M
•
10
•
1
lime-nlp/safer-instruct
Viewer
•
Updated
•
11.2k
•
28
•
1