Introduces the MMLU benchmark of 57 tasks and shows that current models, including GPT-3, achieve low accuracy far below expert level across academic and professional domains.
hub Canonical reference
ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
Canonical reference. 83% of citing Pith papers cite this work as background.
60
Pith papers citing it
Background
83% of classified citations
abstract
Increasing model size when pretraining natural language representations often results in improved performance on downstream tasks. However, at some point further model increases become harder due to GPU/TPU memory limitations and longer training times. To address these problems, we present two parameter-reduction techniques to lower memory consumption and increase the training speed of BERT. Comprehensive empirical evidence shows that our proposed methods lead to models that scale much better compared to the original BERT. We also use a self-supervised loss that focuses on modeling inter-sentence coherence, and show it consistently helps downstream tasks with multi-sentence inputs. As a result, our best model establishes new state-of-the-art results on the GLUE, RACE, and \squad benchmarks while having fewer parameters compared to BERT-large. The code and the pretrained models are available at https://github.com/google-research/ALBERT.
hub tools
citation-role summary
background 5
other 1
citation-polarity summary
representative citing papers
GPT-3 shows that scaling an autoregressive language model to 175 billion parameters enables strong few-shot performance across diverse NLP tasks via in-context prompting without fine-tuning.
ELECTRA replaces masked language modeling with replaced token detection, yielding contextual representations that outperform BERT at equal compute and match larger models like RoBERTa with far less compute.
Embedding norms in contrastive models encode semantic properties via optimization dynamics under scale-invariant losses.
LoopMoE is a looped MoE language model that outperforms matched vanilla MoE on 8 of 9 downstream benchmarks at 3B scale and continues to outperform at 9B scale under strictly controlled budgets.
RISE is an inference-time semantic reranking framework that refines low-confidence predictions in rhetorical role labeling using contrastively learned label representations, delivering an average +9.15 macro-F1 gain on hard examples across eight datasets and seven models.
LoopQ provides a loop-aware PTQ framework for recursive Transformers that mitigates distribution shift, state reuse, and recursive error accumulation, yielding 68.8% higher average accuracy and 87.7% lower perplexity under W4A4 versus static baselines.
NodePFN pre-trains on synthetic graphs with controllable homophily and causal feature-label models to achieve 71.27 average accuracy on 23 node classification benchmarks without graph-specific training.
Open-source LLMs detect phishing intent at high rates but still generate actionable phishing content, and GuardPhish supplies a dataset plus modular classifiers to close the gap.
SecureRouter accelerates secure transformer inference by 1.95x via an encrypted router that selects input-adaptive models from an MPC-optimized pool with negligible accuracy loss.
LA-Sign achieves state-of-the-art skeleton-based sign language recognition on WLASL and MSASL by using recurrent looped transformers with adaptive hyperbolic geometry alignment.
SHREC is a new benchmark dataset of embodied human-robot conversations that shows substantial performance gaps in state-of-the-art foundation models on tasks involving social error detection and rationale generation.
A recurrent-depth architecture enables language models to improve reasoning performance by iterating computation in latent space, achieving gains equivalent to much larger models on benchmarks.
BART introduces a denoising pretraining method for seq2seq models that matches RoBERTa on GLUE and SQuAD while setting new state-of-the-art results on abstractive summarization, dialogue, and QA with up to 6 ROUGE gains.
T5 casts all NLP tasks as text-to-text generation, systematically explores pre-training choices, and reaches strong performance on summarization, QA, classification and other tasks via large-scale training on the Colossal Clean Crawled Corpus.
Intra-layer model parallelism in PyTorch enables training of 8.3B-parameter transformers, achieving SOTA perplexity of 10.8 on WikiText103 and 66.5% accuracy on LAMBADA.
XLNet is a generalized autoregressive pretraining method that learns bidirectional contexts via permutation-based factorization and outperforms BERT on 20 NLP tasks.
A 2x2 ablation shows repeated shared access enables grokking while addressable memory (not recurrence) enables edit propagation in transformer variants on synthetic KG QA.
LARM enables test-time compute scaling in non-autoregressive ASR via depth-conditioned looping with CTC checkpoints, supervision embeddings, FiLM conditioning, and posterior feedback, yielding lower WER on LibriSpeech with more loops.
Q-K=V projection sharing in transformers matches standard QKV performance with 50% KV cache reduction and combines with GQA/MQA for up to 96.9% reduction across vision and language tasks.
Graph-GRPO builds a dependency graph over CoT steps and propagates outcome rewards to enable finer credit assignment in generative relevance modeling for e-commerce search.
Explicitly disentangling semantic and positional streams in a Transformer encoder reveals that absolute positional representations collapse to a 2D document-structure manifold, attention heads specialize by role, and the approach improves linguistic probing performance on 49 of 65 phenomena.
LoopMDM loops early-middle layers in masked diffusion models to match same-size MDM performance with up to 3.3x fewer training FLOPs and outperform on reasoning tasks by up to 8.5 points on GSM8K.
GRAM is a latent-variable generative model that performs recursive reasoning via stochastic trajectories, trained with amortized variational inference to support multi-hypothesis reasoning and unconditional generation.
citing papers explorer
-
Measuring Massive Multitask Language Understanding
Introduces the MMLU benchmark of 57 tasks and shows that current models, including GPT-3, achieve low accuracy far below expert level across academic and professional domains.
-
Language Models are Few-Shot Learners
GPT-3 shows that scaling an autoregressive language model to 175 billion parameters enables strong few-shot performance across diverse NLP tasks via in-context prompting without fine-tuning.
-
ELECTRA: Pre-training Text Encoders as Discriminators Rather Than Generators
ELECTRA replaces masked language modeling with replaced token detection, yielding contextual representations that outperform BERT at equal compute and match larger models like RoBERTa with far less compute.
-
Optimization Dynamics Imprint Semantic Specificity in Contrastive Embedding Norms
Embedding norms in contrastive models encode semantic properties via optimization dynamics under scale-invariant losses.
-
LoopMoE: Unifying Iterative Computation with Mixture-of-Experts for Language Modeling
LoopMoE is a looped MoE language model that outperforms matched vanilla MoE on 8 of 9 downstream benchmarks at 3B scale and continues to outperform at 9B scale under strictly controlled budgets.
-
Semantic Reranking at Inference Time for Hard Examples in Rhetorical Role Labeling
RISE is an inference-time semantic reranking framework that refines low-confidence predictions in rhetorical role labeling using contrastively learned label representations, delivering an average +9.15 macro-F1 gain on hard examples across eight datasets and seven models.
-
LoopQ: Quantization for Recursive Transformers
LoopQ provides a loop-aware PTQ framework for recursive Transformers that mitigates distribution shift, state reuse, and recursive error accumulation, yielding 68.8% higher average accuracy and 87.7% lower perplexity under W4A4 versus static baselines.
-
Learning Posterior Predictive Distributions for Node Classification from Synthetic Graph Priors
NodePFN pre-trains on synthetic graphs with controllable homophily and causal feature-label models to achieve 71.27 average accuracy on 23 node classification benchmarks without graph-specific training.
-
GuardPhish: Securing Open-Source LLMs from Phishing Abuse
Open-source LLMs detect phishing intent at high rates but still generate actionable phishing content, and GuardPhish supplies a dataset plus modular classifiers to close the gap.
-
SecureRouter: Encrypted Routing for Efficient Secure Inference
SecureRouter accelerates secure transformer inference by 1.95x via an encrypted router that selects input-adaptive models from an MPC-optimized pool with negligible accuracy loss.
-
LA-Sign: Looped Transformers with Geometry-aware Alignment for Skeleton-based Sign Language Recognition
LA-Sign achieves state-of-the-art skeleton-based sign language recognition on WLASL and MSASL by using recurrent looped transformers with adaptive hyperbolic geometry alignment.
-
Social Human Robot Embodied Conversation (SHREC) Dataset: Benchmarking Foundational Models' Social Reasoning
SHREC is a new benchmark dataset of embodied human-robot conversations that shows substantial performance gaps in state-of-the-art foundation models on tasks involving social error detection and rationale generation.
-
Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach
A recurrent-depth architecture enables language models to improve reasoning performance by iterating computation in latent space, achieving gains equivalent to much larger models on benchmarks.
-
BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension
BART introduces a denoising pretraining method for seq2seq models that matches RoBERTa on GLUE and SQuAD while setting new state-of-the-art results on abstractive summarization, dialogue, and QA with up to 6 ROUGE gains.
-
Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
T5 casts all NLP tasks as text-to-text generation, systematically explores pre-training choices, and reaches strong performance on summarization, QA, classification and other tasks via large-scale training on the Colossal Clean Crawled Corpus.
-
Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism
Intra-layer model parallelism in PyTorch enables training of 8.3B-parameter transformers, achieving SOTA perplexity of 10.8 on WikiText103 and 66.5% accuracy on LAMBADA.
-
XLNet: Generalized Autoregressive Pretraining for Language Understanding
XLNet is a generalized autoregressive pretraining method that learns bidirectional contexts via permutation-based factorization and outperforms BERT on 20 NLP tasks.
-
Repeated Shared Access Enables Grokking, but Edit Propagation Depends on an Addressable Memory
A 2x2 ablation shows repeated shared access enables grokking while addressable memory (not recurrence) enables edit propagation in transformer variants on synthetic KG QA.
-
Test-Time Compute Scaling for ASR with Depth-Conditioned Looped Transformers
LARM enables test-time compute scaling in non-autoregressive ASR via depth-conditioned looping with CTC checkpoints, supervision embeddings, FiLM conditioning, and posterior feedback, yielding lower WER on LibriSpeech with more loops.
-
Do Transformers Need Three Projections? Systematic Study of QKV Variants
Q-K=V projection sharing in transformers matches standard QKV performance with 50% KV cache reduction and combines with GQA/MQA for up to 96.9% reduction across vision and language tasks.
-
Graph-GRPO: Dependency-Aware Credit Assignment for Generative E-commerce Search Relevance
Graph-GRPO builds a dependency graph over CoT steps and propagates outcome rewards to enable finer credit assignment in generative relevance modeling for e-commerce search.
-
Give it Space! Explicit Disentangling of Positional and Semantic Representations in Encoders
Explicitly disentangling semantic and positional streams in a Transformer encoder reveals that absolute positional representations collapse to a 2D document-structure manifold, attention heads specialize by role, and the approach improves linguistic probing performance on 49 of 65 phenomena.
-
Looped Diffusion Language Models
LoopMDM loops early-middle layers in masked diffusion models to match same-size MDM performance with up to 3.3x fewer training FLOPs and outperform on reasoning tasks by up to 8.5 points on GSM8K.
-
Generative Recursive Reasoning
GRAM is a latent-variable generative model that performs recursive reasoning via stochastic trajectories, trained with amortized variational inference to support multi-hypothesis reasoning and unconditional generation.
-
BoolXLLM: LLM-Assisted Explainability for Boolean Models
BoolXLLM augments an existing Boolean rule learner with LLMs for feature selection, discretization thresholds, and natural-language rule translation to improve interpretability while preserving accuracy.
-
Shortcut Solutions Learned by Transformers Impair Continual Compositional Reasoning
BERT learns shortcut solutions that impair generalization and forward transfer in continual LEGO, while ALBERT learns loop-like solutions for better performance, yet both fail at cross-experience composition, with ALBERT rescued by mixed-data training.
-
Fight Poison with Poison: Enhancing Robustness in Few-shot Machine-Generated Text Detection with Adversarial Training
REACT uses a RAG-powered attacker to generate challenging adversarial examples and trains a detector with contrastive learning in an alternating loop, raising average F1 by 4.95 points and lowering attack success rate by 3.66 points across tested settings.
-
ADE: Adaptive Dictionary Embeddings -- Scaling Multi-Anchor Representations to Large Language Models
ADE scales multi-anchor word representations to transformers via Vocabulary Projection, Grouped Positional Encoding, and context-aware reweighting, achieving 98.7% fewer trainable parameters than DeBERTa-v3-base while matching or exceeding it on two text-classification benchmarks and compressing the
-
RedNote-Vibe: A Dataset for Capturing Temporal Dynamics of AI-Generated Text in Lifestyle Social Media
RedNote-Vibe supplies a longitudinal dataset of AI versus human lifestyle posts from 2020 to mid-2025 plus the PLAD detection framework that applies cognitive psychology signatures for improved AI-text identification.
-
Atlas: Few-shot Learning with Retrieval Augmented Language Models
Atlas reaches over 42% accuracy on Natural Questions with only 64 examples, outperforming a 540B-parameter model by 3% with 50x fewer parameters.
-
Language Models (Mostly) Know What They Know
Language models show good calibration when asked to estimate the probability that their own answers are correct, with performance improving as models get larger.
-
Unsupervised Dense Information Retrieval with Contrastive Learning
Contrastive learning trains unsupervised dense retrievers that beat BM25 on most BEIR datasets and support cross-lingual retrieval across scripts.
-
A General Language Assistant as a Laboratory for Alignment
Ranked preference modeling outperforms imitation learning for language model alignment and scales more favorably with model size.
-
Scaling Laws for Transfer
Effective data transferred from pre-training to fine-tuning is described by a power law in model parameter count and fine-tuning dataset size, acting like a multiplier on the fine-tuning data.
-
Aligning AI With Shared Human Values
Introduces ETHICS benchmark showing current language models have promising but incomplete ability to predict basic human ethical judgments on text scenarios.
-
How Much Knowledge Can You Pack Into the Parameters of a Language Model?
Fine-tuned language models store knowledge in parameters to answer questions competitively with retrieval-based open-domain QA systems.
-
HuggingFace's Transformers: State-of-the-art Natural Language Processing
Hugging Face releases an open-source Python library that supplies a unified API and pretrained weights for major Transformer architectures used in natural language processing.
-
BetXplain: An Explanation-Annotated Dataset for Detecting Manipulative Betting Advertisements on Social Media
Introduces BetXplain, an explanation-annotated dataset of social media betting ads collected from Instagram and Reddit for detecting manipulative and deceptive advertising.
-
From Sentiment to Actionable Insights: Public Sentiment Analysis of Advanced Air Mobility
ModernBERT-labeled analysis of 306k social posts yields six major public-sentiment clusters on Advanced Air Mobility spanning noise, safety, regulation, workforce, drones, and military use.
-
Multi-Bitwidth Quantization for LLMs Using Additive Codebooks
Drop-by-Drop uses additive codebooks and Matryoshka-style training to produce one LLM model whose ordered codebook subsets give accurate reconstructions at successively higher bitwidths under a weighted MSE distortion.
-
Rethinking Depth: A study of the Recursive-Transformer for Speech Recognition
Experimental study finds Recursive-Transformer for ASR encoders achieves comparable performance with 66% fewer parameters when limited recursion is applied in the latent space.
-
Artificial Intelligence for Mathematical Reasoning: An Integrated Survey of Language Models, Neuro-symbolic Systems, and Verified Discovery
AI for mathematics is best described as a supervision ladder — final answers, programs, process rewards, proof-assistant kernels — culminating in verified-discovery workflows.
-
When Meaning Travels: A Granular Lens on Hybrid-MoE's Role in Idiomatic Understanding for Language Models
HybridMoE with controlled hybridization and idiomatic property signals yields 5-6% gains in figurative language representation for multilingual vision-language models.
-
BitTP: The Lightweight Trajectory Prediction Model with BitLLM for Edge-Devices
BitTP applies weight-only 1.58-bit quantization to LLM trajectory predictors, claiming improved ADE/FDE over BF16 baseline with reduced resource demands on edge devices.
-
No Free Swap: Protocol-Dependent Layer Redundancy in Transformers
Replacement and interchange swap-KL protocols for layer redundancy in transformers disagree on pruning safety, with the gap growing during training on Pythia models and producing different removal costs on Qwen3-8B versus Llama-3.1-8B.
-
SMolLM: Small Language Models Learn Small Molecular Grammar
A 53K-parameter weight-shared transformer generates novel valid SMILES at 95% rate on ZINC-250K and resolves constraints hierarchically via bracket, ring, and valence stages as shown by probing and ablation.
-
Hyperloop Transformers
A looped Transformer with matrix-valued hyper-connections matches depth-matched baselines at ~50% fewer parameters, including under post-training quantization.
-
Scaling Latent Reasoning via Looped Language Models
A 1.4B and a 2.6B looped (weight-tied, recurrent-depth) language model trained on 7.7T tokens match or exceed several 4B–8B transformer baselines on selected reasoning benchmarks.
-
Model Compression vs. Adversarial Robustness: An Empirical Study on Language Models for Code
Empirical tests show compressed code language models retain task performance but suffer markedly lower robustness under four standard adversarial attacks.
-
MegaFake: A Theory-Driven Dataset of Fake News Generated by Large Language Models
Authors create LLM-Fake Theory integrating social psychology, then use a prompt engineering pipeline to build the MegaFake dataset of LLM-generated fake news for advancing detection methods.