List your company

Intent Classification on MASSIVE (test)

89.47In-Scope AccuracyFT Teacher
67.765273.400179.03584.6699May 11, 2023Oct 26, 2023Apr 11, 2024Sep 26, 2024Mar 13, 2025Aug 28, 2025Feb 13, 2026

Evaluation results

MethodLinks
FT TeacherModel Role=Teacher2026.0289.47———
CUDStudent Model Architecture=6L-768D2026.0289.44———
AD-KDStudent Model Architecture=6L-768D2026.0289.3———
CKDStudent Model Architecture=6L-768D2026.0289.3———
LKDStudent Model Architecture=6L-768D2026.0288.94———
TinyBERTStudent Model Architecture=6L-768D, trained with distillation at pretraining stage=true2026.0288.84———
MGSKDStudent Model Architecture=6L-768D, trained with distillation at pretraining stage=true2026.0288.56———
PKDStudent Model Architecture=6L-768D2026.0288.17———
PKDStudent Model Architecture=4L-256D2026.0287.23———
CUDStudent Model Architecture=4L-256D2026.0287.22———
CKDStudent Model Architecture=4L-256D2026.0286.6———
AD-KDStudent Model Architecture=4L-256D2026.0284.17———
LKDStudent Model Architecture=4L-256D2026.0282———
GPT-3LLM Intents=60, parameters=175B, Evaluation Mode=zero-shot2023.0573.9———
Flan-T5-XXLLLM Intents=60, parameters=11B, Evaluation Mode=zero-shot2023.0573.3———
GPT-3LLM Intents=5, parameters=175B, Evaluation Mode=zero-shot2023.0569.2———
Flan-T5-XXLLLM Intents=5, parameters=11B, Evaluation Mode=zero-shot2023.0568.6———
OLABackbone=Qwen3, MLP Width=256, Activation Scheme=OLA2026.05——85.56—
PreciseBackbone=Qwen3, MLP Width=256, Activation Scheme=Prec.2026.05——85.56—
QUAD4FHEBackbone=Qwen3, MLP Width=256, Activation Scheme=Quad.2026.05——85.4598.69
ReLUBackbone=Qwen3, MLP Width=256, Activation Scheme=ReLU2026.05——85.56—
Remez-7Backbone=Qwen3, MLP Width=256, Activation Scheme=Rmz-72026.05——85.49—
SquareBackbone=Qwen3, MLP Width=256, Activation Scheme=Sq.2026.05——52.4—