AI Paper Atlas

Research map / Benchmarks / PTB

PTB leaderboard

35 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on PTB. Best perplexity: 18.78 (Agile-Quant-8, AAAI 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.

PTB · perplexity

35 papers, each with its best reported number for its own method; lower is better.

#perplexity ↓MethodPaperTopic
118.78Agile-Quant-8
OPT-1.3B
Agile-Quant: Activation-Guided Quantization for Faster Inference of LLMs on the Edge
AAAI 2024 · Table: OPT and BLOOM results o
Quantization
222.87BESA
LLaMA2-70B
BESA: Pruning Large Language Models with Blockwise Parameter-Efficient Sparsity Allocation
ICLR 2024 · Table 1
Pruning and efficient tuning
327.06MeRino-64MMeRino: Entropy-Driven Design for Generative Language Models on IoT Devices
AAAI 2025 · Table: Performance comparison
AI and society
434.9BigMac
GPT3-XL
BigMac: A Communication-Efficient Mixture-of-Experts Model Structure for Fast Training and Inference
AAAI 2025 · Table: downstream results afte
Mixture of experts
541.15AdaFisherW
scaled-down GPT-1
AdaFisher: Adaptive Second Order Optimization via Fisher Information
ICLR 2025 · Table: Language Modeling perfo
Optimizers
644.6Post-RBN
Post-Norm Transformer
Understanding the Failure of Batch Normalization for Transformers in NLP
NeurIPS 2022 · Table: Results for performance
RNNs and normalization
746.01AWD-LSTM + MoS + Partial Shuffled + Ours
AWD-LSTM + MoS + Partial Shuffled
Improving Neural Language Modeling via Adversarial Training
ICML 2019 · Table 1
Attacks on text and graphs
846.54AWD-LSTM-MoS + dynamic evaluation with FRAGE
AWD-LSTM-MoS + dynamic evaluation
FRAGE: Frequency-Agnostic Word Representation
NeurIPS 2018 · Table 2
Recognition and long-tail
952.34AWD-LSTM-DOC + char3-MS-vec
AWD-LSTM-DOC
Character n-Gram Embeddings to Improve RNN Language Models
AAAI 2019 · Perplexities of the proposed m
Language, safety and interpretability
1053.31LoT (AWD-LSTM)
AWD-LSTM
Learning from Teaching Regularization: Generalizable Correlations Should be Easy to Imitate
NeurIPS 2024 · Table 1
Recognition and long-tail
25 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.