AI Paper Atlas

Research map / Benchmarks / SST-2

SST-2 leaderboard

148 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on SST-2. Best accuracy: 98.16 (Vector-ICL (finetuned), ICLR 2025).

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.

SST-2 · accuracy

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

#accuracy ↑MethodPaperTopic
198.16Vector-ICL (finetuned)
Llama-3.1-8B; NV-Embed-v1
Vector-ICL: In-context Learning with Continuous Vector Representations
ICLR 2025 · Table: sentiment analysis and
Language, safety and interpretability
297.08OFA
Llama3-8B-Instruct
Optimization Inspired Few-Shot Adaptation for Large Language Models
NeurIPS 2025 · Table 1
Instruction and fine-tuning
397RoBERTa Large + AT + CTR
RoBERTa Large
Improved Text Classification via Contrastive Adversarial Training
AAAI 2022 · Table 2
Certified and empirical robustness
496.4PromptQuine (4-shot ICL)
Meta-Llama-3-8B-Instruct
Evolving Prompts In-Context: An Open-ended, Self-replicating Perspective
ICML 2025 · Table: classification results
Language, safety and interpretability
596.38CB-LLMs
Llama3-8B
Concept Bottleneck Large Language Models
ICLR 2025 · Table: Accuracy, steerability,
Language, safety and interpretability
696.2SIFT
RoBERTa-Large
Sparse is Enough in Fine-tuning Pre-trained Large Language Models
ICML 2024 · Table 1
Recognition and long-tail
796.1PortLLM
Mistral-7B
PortLLM: Personalizing Evolving Large Language Models with Training-Free and Portable Model Patches
ICLR 2025 · Table: Zero-shot performance c
Instruction and fine-tuning
896.01DePT
Llama-2-13B
DePT: Decomposed Prompt Tuning for Parameter-Efficient Fine-tuning
ICLR 2024 · Appendix table: Llama-2 result
Pruning and efficient tuning
996CL-Attack
Qwen2-7B
CL-Attack: Textual Backdoor Attacks via Cross-Lingual Triggers
AAAI 2025 · Backdoor attack results table
Backdoor and poisoning
1096FedSA-LoRA
RoBERTa-large
Selective Aggregation for Low-Rank Adaptation in Federated Learning
ICLR 2025 · GLUE results table
Personalized and heterogeneous FL
126 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

SST-2 · accuracy (dev)

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

#accuracy ↑MethodPaperTopic
197.1R4F
RoBERTa-Large
Better Fine-Tuning by Reducing Representational Collapse
ICLR 2021 · GLUE results table
Pruning and efficient tuning
296.4i-GELU
RoBERTa-Large
I-BERT: Integer-only BERT Quantization
ICML 2021 · Table 3
Quantization
395.2ALBERT-xxlarge
ALBERT
ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
ICLR 2020 · Table: Overall Comparison betw
Attention and transformers
494.2MAM Adapter
RoBERTa-BASE
Towards a Unified View of Parameter-Efficient Transfer Learning
ICLR 2022 · Table: Accuracy on the dev set
Pruning and efficient tuning
594.2gMLP-basePay Attention to MLPs
NeurIPS 2021 · Table 6
Attention and transformers
693.3ROSITA
BERT_BASE
ROSITA: Refined BERT cOmpreSsion with InTegrAted techniques
AAAI 2021 · Comparison with Existing Metho
Pruning and efficient tuning
793.2UniLMv2 baseUniLMv2: Pseudo-Masked Language Models for Unified Language Model Pre-Training
ICML 2020 · Table: Comparisons between the
Language, safety and interpretability
892.66Q-BERT 4-bit, 128 groups
BERTBASE
Q-BERT: Hessian Based Ultra Low Precision Quantization of BERT
AAAI 2020 · Table 3
Quantization
992.31ReLU-SNNK-Adapter
BERT-base
Scalable Neural Network Kernels
ICLR 2024 · GLUE results table
Pruning and efficient tuning
1092.3YOSO-E
BERT-base
You Only Sample (Almost) Once: Linear Cost Self-Attention Via Bernoulli Sampling
ICML 2021 · Table (BERT-base MLM and GLUE)
Attention and transformers
4 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.