A framework to identify and convert foldable layer normalizations to RMSNorm for exact equivalence and faster inference in deep neural networks.
hub
Journal of machine learning research , volume=
28 Pith papers cite this work. Polarity classification is still indexing.
28
Pith papers citing it
hub tools
representative citing papers
ALiBi bias is the expectation of positional LSH-induced block masks, yielding spectral and max-norm approximation bounds that reduce long-context biased attention to randomized short-context unbiased attention.
Code LLMs generate substantially worse comments outside English, and no tested automatic metric or LLM judge reliably matches human assessment of those outputs.
TCRTransBench provides a new benchmark with bidirectional TCR-peptide generation tasks, a large validated dataset, and metrics to evaluate neural models for immunological sequence modeling.
PRISM lets pre-trained text-to-image models handle long prompts by breaking them into compositional parts, predicting noise separately, and merging outputs via energy-based conjunction, matching fine-tuned models while generalizing better to prompts over 500 tokens.
Theoretical analysis of continual factual knowledge acquisition shows data replay stabilizes pretrained knowledge by shifting convergence dynamics while regularization only slows forgetting, leading to the STOC method for attention-based replay selection.
Synthetic pre-pre-training on structured data improves LLM robustness to noisy pre-training, matching baseline loss with up to 49% fewer natural tokens for a 1B model.
ZAYA1-8B is a reasoning MoE model with 700M active parameters that matches larger models on math and coding benchmarks and reaches 91.9% on AIME'25 via Markovian RSA test-time compute.
GEM achieves 65.19% joint goal accuracy on MultiWOZ 2.2 by routing between a graph neural network expert for dialogue structure and a T5 expert for sequences, plus ReAct agents for value generation, outperforming prior SOTA methods.
Verbal-R3 uses a verbal reranker to generate analytic narratives that guide retrieval and reasoning in LLMs, achieving SOTA results on complex QA benchmarks.
Activation steering is cast as constrained optimization that minimizes collateral damage by weighting perturbations according to the empirical second-moment matrix of activations instead of assuming isotropy.
Recurrent Transformers add per-layer recurrent memory via self-attention on own activations plus a tiling algorithm that reduces training memory traffic, yielding better C4 pretraining cross-entropy than parameter-matched standard transformers with fewer layers.
LBLLM achieves better accuracy than prior binarization methods for LLMs by decoupling weight and activation quantization through initialization, layer-wise distillation, and learnable activation scaling.
LLMs show mixed results on authorship verification, post generation, and attribute inference from Twitter data, with new frameworks and user studies establishing benchmarks for these analytics tasks.
TLoRA jointly optimizes LoRA initialization via task-data SVD and sensitivity-driven rank allocation, delivering stronger results than standard LoRA across NLU, reasoning, math, code, and chat tasks while using fewer trainable parameters.
REGLU guides LoRA-based unlearning via representation subspaces and orthogonal regularization to outperform prior methods on forget-retain trade-off in LLM benchmarks.
CogVideoX generates coherent 10-second text-to-video outputs at high resolution using a 3D VAE, expert adaptive LayerNorm transformer, progressive training, and a custom data pipeline, claiming state-of-the-art results.
NV-Embed achieves first place on the MTEB leaderboard across 56 tasks by combining a latent attention layer, causal-mask removal, two-stage contrastive training, and data curation for LLM-based embedding models.
Cosine similarity poorly predicts performance degradation from layer removal in LLMs, making direct accuracy-drop ablation a more reliable relevance metric.
TaTok is a theoretically grounded adaptive tokenization method that uses global tokens and cumulative conditional entropy filtering to reduce redundancy while improving reconstruction quality over fixed-rate patch tokenization.
SmolLM2 is a 1.7B-parameter language model that outperforms Qwen2.5-1.5B and Llama3.2-1B after overtraining on 11 trillion tokens using custom FineMath, Stack-Edu, and SmolTalk datasets in a multi-stage pipeline.
ModernBERT is a new bidirectional encoder model achieving SOTA performance on diverse classification and retrieval benchmarks while offering superior speed and memory efficiency for long-context inference.
ClaHF converts instance labels into preference signals via candidate predictions and a reward model, then applies RL optimization to improve text classification accuracy and calibration.
Two-stage multilingual then dataset-specific adapter fine-tuning of Gemma-3-27b with headword XML mention representation and iterative annotation achieved first place in the CRAC 2026 LLM track.
citing papers explorer
-
Enjoy Your Layer Normalization with the Computational Efficiency of RMSNorm
A framework to identify and convert foldable layer normalizations to RMSNorm for exact equivalence and faster inference in deep neural networks.
-
Positional LSH: Binary Block Matrix Approximation for Attention with Linear Biases
ALiBi bias is the expectation of positional LSH-induced block masks, yielding spectral and max-norm approximation bounds that reduce long-context biased attention to randomized short-context unbiased attention.
-
Evaluating Non-English Developer Support in Machine Learning for Software Engineering
Code LLMs generate substantially worse comments outside English, and no tested automatic metric or LLM judge reliably matches human assessment of those outputs.
-
TCRTransBench: A Comprehensive Benchmark for Bidirectional TCR-Peptide Sequence Generation
TCRTransBench provides a new benchmark with bidirectional TCR-peptide generation tasks, a large validated dataset, and metrics to evaluate neural models for immunological sequence modeling.
-
Long-Text-to-Image Generation via Compositional Prompt Decomposition
PRISM lets pre-trained text-to-image models handle long prompts by breaking them into compositional parts, predicting noise separately, and merging outputs via energy-based conjunction, matching fine-tuned models while generalizing better to prompts over 500 tokens.
-
Towards Understanding Continual Factual Knowledge Acquisition of Language Models: From Theory to Algorithm
Theoretical analysis of continual factual knowledge acquisition shows data replay stabilizes pretrained knowledge by shifting convergence dynamics while regularization only slows forgetting, leading to the STOC method for attention-based replay selection.
-
Synthetic Pre-Pre-Training Improves Language Model Robustness to Noisy Pre-Training Data
Synthetic pre-pre-training on structured data improves LLM robustness to noisy pre-training, matching baseline loss with up to 49% fewer natural tokens for a 1B model.
-
ZAYA1-8B Technical Report
ZAYA1-8B is a reasoning MoE model with 700M active parameters that matches larger models on math and coding benchmarks and reaches 91.9% on AIME'25 via Markovian RSA test-time compute.
-
GEM: Graph-Enhanced Mixture-of-Experts with ReAct Agents for Dialogue State Tracking
GEM achieves 65.19% joint goal accuracy on MultiWOZ 2.2 by routing between a graph neural network expert for dialogue structure and a T5 expert for sequences, plus ReAct agents for value generation, outperforming prior SOTA methods.
-
Verbal-R3: Verbal Reranker as the Missing Bridge between Retrieval and Reasoning
Verbal-R3 uses a verbal reranker to generate analytic narratives that guide retrieval and reasoning in LLMs, achieving SOTA results on complex QA benchmarks.
-
Minimizing Collateral Damage in Activation Steering
Activation steering is cast as constrained optimization that minimizes collateral damage by weighting perturbations according to the empirical second-moment matrix of activations instead of assuming isotropy.
-
The Recurrent Transformer: Greater Effective Depth and Efficient Decoding
Recurrent Transformers add per-layer recurrent memory via self-attention on own activations plus a tiling algorithm that reduces training memory traffic, yielding better C4 pretraining cross-entropy than parameter-matched standard transformers with fewer layers.
-
LBLLM: Lightweight Binarization of Large Language Models via Three-Stage Distillation
LBLLM achieves better accuracy than prior binarization methods for LLMs by decoupling weight and activation quantization through initialization, layer-wise distillation, and learnable activation scaling.
-
Assessing Capabilities of Large Language Models in Social Media Analytics: A Multi-task Quest
LLMs show mixed results on authorship verification, post generation, and attribute inference from Twitter data, with new frameworks and user studies establishing benchmarks for these analytics tasks.
-
TLoRA: Task-aware Low Rank Adaptation of Large Language Models
TLoRA jointly optimizes LoRA initialization via task-data SVD and sensitivity-driven rank allocation, delivering stronger results than standard LoRA across NLU, reasoning, math, code, and chat tasks while using fewer trainable parameters.
-
Representation-Guided Parameter-Efficient LLM Unlearning
REGLU guides LoRA-based unlearning via representation subspaces and orthogonal regularization to outperform prior methods on forget-retain trade-off in LLM benchmarks.
-
CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer
CogVideoX generates coherent 10-second text-to-video outputs at high resolution using a 3D VAE, expert adaptive LayerNorm transformer, progressive training, and a custom data pipeline, claiming state-of-the-art results.
-
NV-Embed: Improved Techniques for Training LLMs as Generalist Embedding Models
NV-Embed achieves first place on the MTEB leaderboard across 56 tasks by combining a latent attention layer, causal-mask removal, two-stage contrastive training, and data curation for LLM-based embedding models.
-
Rethinking Layer Relevance in Large Language Models Beyond Cosine Similarity
Cosine similarity poorly predicts performance degradation from layer removal in LLMs, making direct accuracy-drop ablation a more reliable relevance metric.
-
Mutual Enhancement Between Global Tokens and Patch Tokens: From Theory to Practice
TaTok is a theoretically grounded adaptive tokenization method that uses global tokens and cumulative conditional entropy filtering to reduce redundancy while improving reconstruction quality over fixed-rate patch tokenization.
-
SmolLM2: When Smol Goes Big -- Data-Centric Training of a Small Language Model
SmolLM2 is a 1.7B-parameter language model that outperforms Qwen2.5-1.5B and Llama3.2-1B after overtraining on 11 trillion tokens using custom FineMath, Stack-Edu, and SmolTalk datasets in a multi-stage pipeline.
-
Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference
ModernBERT is a new bidirectional encoder model achieving SOTA performance on diverse classification and retrieval benchmarks while offering superior speed and memory efficiency for long-context inference.
-
ClaHF: A Human Feedback-inspired Reinforcement Learning Framework for Improving Classification Tasks
ClaHF converts instance labels into preference signals via candidate predictions and a reward model, then applies RL optimization to improve text classification accuracy and calibration.
-
Closing the Gap at CRAC 2026: Two-Stage Adaptation for LLM-Based Multilingual Coreference Resolution
Two-stage multilingual then dataset-specific adapter fine-tuning of Gemma-3-27b with headword XML mention representation and iterative annotation achieved first place in the CRAC 2026 LLM track.
-
Enhancing Target-Guided Proactive Dialogue Systems via Conversational Scenario Modeling and Intent-Keyword Bridging
Conversational scenario modeling from user profiles and domain knowledge, combined with intent-keyword bridging, improves proactivity, fluency, and informativeness in target-guided proactive dialogue systems.
-
Step-Video-T2V Technical Report: The Practice, Challenges, and Future of Video Foundation Model
Step-Video-T2V describes a 30B-parameter text-to-video model with custom Video-VAE, 3D DiT, flow matching, and Video-DPO that claims state-of-the-art results on a new internal benchmark.
- Diagnosing and Correcting Concept Omission in Multimodal Diffusion Transformers
- Block-wise Codeword Embedding for Reliable Multi-bit Text Watermarking