AI Paper Atlas

Research map / Benchmarks / MNLI

MNLI leaderboard

49 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on MNLI. Best accuracy: 90.69 (LoRA init A, NeurIPS 2024).

Numbers are as each paper reports them, read by AI from its results tables and not checked by hand. Papers can differ in setting (training data, model size, extra data); the topic column shows what each paper is about. Open the paper before you compare or quote a number.

MNLI · accuracy

49 papers, each with its best reported number for its own method; higher is better.

#accuracy ↑MethodPaperTopic
190.69LoRA init A
RoBERTa-Large
The Impact of Initialization on LoRA Finetuning Dynamics
NeurIPS 2024 · text
Sample complexity
290.6SIFT
RoBERTa-Large
Sparse is Enough in Fine-tuning Pre-trained Large Language Models
ICML 2024 · Table 1
Recognition and long-tail
390.34LoRAM
DeBERTa-v3-base
The Primacy of Magnitude in Low-Rank Adaptation
NeurIPS 2025 · GLUE results table
Pruning and efficient tuning
487.86BOMF
RoBERTa-base
Model Fusion through Bayesian Optimization in Language Model Fine-Tuning
NeurIPS 2024 · Table 1
Bayesian optimization
587.8DP LoRA
RoBERTa-Large
Differentially Private Fine-tuning of Language Models
ICLR 2022 · Table: RoBERTa-Large accuracy
Differential privacy
687.4DiSP
Qwen2.5-7B
Easier to Judge than to Find: Predicting In-Context Learning Success for Demonstration Selection
ICML 2026 · Table 1
Language, safety and interpretability
787.34SUMO (SVD, rank=4)
RoBERTa-Base
SUMO: Subspace-Aware Moment-Orthogonalization for Accelerating Memory-Efficient LLM Training
NeurIPS 2025 · GLUE results table
Optimizers
887.21Gradual Stacking
BERT-LARGE
Efficient Training of Language Models using Few-Shot Learning
ICML 2023 · Table 8
Language, safety and interpretability
987.04LoRA rank 8, α = 2r
RoBERTa-base
LoRA vs Full Fine-tuning: An Illusion of Equivalence
NeurIPS 2025 · Appendix, Model Accuracies, α
Pruning and efficient tuning
1086.61GACT Adapt 4bit (L1), Bert-large
Bert-large
GACT: Activation Compressed Training for Generic Network Architectures
ICML 2022 · Table: NLP and Graph accuracy
RNNs and normalization
39 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

Not affiliated with any conference, journal or the benchmark's authors. Corrections: contact@latenticlabs.com.