The upper-tail accumulation scale derived from the gap-counting function N_n sets the critical inverse temperature for softmax attention concentration, unifying prior conflicting laws as special cases of different N_n.
super hub Mixed citations
RoFormer: Enhanced Transformer with Rotary Position Embedding
Mixed citation behavior. Most common role is background (46%).
138
Pith papers citing it
Background
46% of classified citations
abstract
Position encoding recently has shown effective in the transformer architecture. It enables valuable supervision for dependency modeling between elements at different positions of the sequence. In this paper, we first investigate various methods to integrate positional information into the learning process of transformer-based language models. Then, we propose a novel method named Rotary Position Embedding(RoPE) to effectively leverage the positional information. Specifically, the proposed RoPE encodes the absolute position with a rotation matrix and meanwhile incorporates the explicit relative position dependency in self-attention formulation. Notably, RoPE enables valuable properties, including the flexibility of sequence length, decaying inter-token dependency with increasing relative distances, and the capability of equipping the linear self-attention with relative position encoding. Finally, we evaluate the enhanced transformer with rotary position embedding, also called RoFormer, on various long text classification benchmark datasets. Our experiments show that it consistently overcomes its alternatives. Furthermore, we provide a theoretical analysis to explain some experimental results. RoFormer is already integrated into Huggingface: \url{https://huggingface.co/docs/transformers/model_doc/roformer}.
hub tools
citation-role summary
background 18
method 8
baseline 1
dataset 1
citation-polarity summary
claims ledger
- abstract Position encoding recently has shown effective in the transformer architecture. It enables valuable supervision for dependency modeling between elements at different positions of the sequence. In this paper, we first investigate various methods to integrate positional information into the learning process of transformer-based language models. Then, we propose a novel method named Rotary Position Embedding(RoPE) to effectively leverage the positional information. Specifically, the proposed RoPE encodes the absolute position with a rotation matrix and meanwhile incorporates the explicit relative
authors
co-cited works
representative citing papers
RULER shows most long-context LMs drop sharply in performance on complex tasks as length and difficulty increase, with only half maintaining results at 32K tokens.
Mamba is a linear-time sequence model using input-dependent selective SSMs that achieves SOTA results across modalities and matches twice-larger Transformers on language modeling with 5x higher inference throughput.
Score entropy loss enables discrete diffusion models (SEDD) that cut perplexity 25-75% versus prior diffusion methods and outperform GPT-2 on language modeling while supporting infilling and compute-quality tradeoffs.
Pythia releases 16 identically trained LLMs with full checkpoints and data tools to study training dynamics, scaling, memorization, and bias in language models.
Introduces the first large-scale GRW dataset for semantic co-speech gesture classification, word recognition, and temporal localization in unconstrained videos, along with benchmarks for the three tasks.
Autoregressive transformers exhibit measurable cognitive fatigue during extended generation, quantified by the Fatigue Index that predicts degradation (AUROC 0.95) and repetition (rho 0.94).
Tensor Cache augments sliding-window attention with an eviction-fed outer-product associative memory and a training correction to improve long-context performance under bounded memory.
POST uses prior-observation adversarial learning on adjacency matrices to reduce spatial over-generalization in graph-based multivariate time series anomaly detection and achieves new SOTA results on detection and channel-wise localization.
CRePE supplies depth-aware positional distributions along curved rays for stable unified-camera control in frozen video DiT models.
Embedding Temporal Logic (ETL) performs runtime monitoring directly in learned embedding spaces using distance-based predicates composed with temporal operators, supported by conformal calibration for reliable predicate evaluation.
Chirality emerges in SMILES translation models through an abrupt encoder-centered reorganization of representations after a long plateau, identified via checkpoint analysis and ablation.
CGAD is a staleness-aware Adam variant for DiLoCo that gates gradients with cosine and exponential decay, proves a convergence bound independent of maximum delay, and demonstrates stable pretraining of 25M to 7B parameter Llama-style models across controlled delays.
Jordan-RoPE realizes a distance-modulated phase basis via non-semisimple Jordan blocks, generating features such as d e^{iωd} for relative positional encoding.
A graph transformer with RL stabilizations is the first to exceed benchmarks for dynamic RMSA, supporting up to 13% more traffic load on networks up to 143 nodes.
A single end-to-end Transformer model unifies stellar labels from heterogeneous spectroscopic surveys into a self-consistent scale without post-hoc recalibration.
Physics-informed transformer with sin^2(theta) encoding, physics-aware positional encoding, multi-task decoder, and three-stage curriculum classifies powder diffraction into 99 extinction groups, with structured errors on symmetry subgroup hierarchy.
A Semantic Progress Function is defined as a 1D curve of cumulative semantic shifts from frame embeddings, supporting a linearization procedure that retimes video sequences for constant-rate semantic evolution.
WildSplatter jointly learns 3D Gaussians and appearance embeddings from unconstrained photo collections to enable fast feed-forward reconstruction and flexible lighting control in 3D Gaussian Splatting.
The work demonstrates masked-token prediction with transformers for model-independent anomaly detection in LHC data, achieving strong results on top-rich BSM signatures like four-top production using VQ-VAE tokenization.
FP32-converged language models enter a post-convergence phase where INT4 quantization error explodes while FP32 perplexity remains stable, with onset tied to fine convergence rather than learning rate decay.
Dual Triangle Attention achieves effective bidirectional attention with built-in positional inductive bias via dual triangular masks, outperforming standard bidirectional attention on position-sensitive tasks and showing strong masked language modeling results with or without positional embeddings.
Activation probes detect hallucinations pre-generation in large LLMs but cannot correct them via steering, with output confidence outperforming on accuracy.
AR-VLA introduces a standalone autoregressive action expert with long-lived memory that generates context-aware continuous actions for VLAs, replacing chunk-based heads with smoother trajectories and maintained task success.
citing papers explorer
-
Hearing to Translate: The Effectiveness of Speech Modality Integration into LLMs
Cascaded systems remain the most reliable for speech translation overall, but recent SpeechLLMs match or outperform them in many conditions while standalone speech models lag.
-
Group Representational Position Encoding
GRAPE unifies RoPE and ALiBi as special cases of group actions on positions, providing a principled design space for positional encodings via SO(d) rotations and GL unipotent transformations.
-
SAM 3: Segment Anything with Concepts
SAM 3 introduces promptable concept segmentation that doubles accuracy of prior systems on images and videos while improving standard SAM segmentation performance.
-
Key and Value Weights Are Probably All You Need: On the Necessity of the Query, Key, Value weight Triplet in Self-Attention Transformers
One of the Q, K or V weights in transformer self-attention is redundant and replaceable by the identity matrix under mild assumptions, reducing parameters by 25 percent with no loss in small-model performance.
-
When Thoughts Meet Facts: Reusable Reasoning for Long-Context LMs
Thought templates derived from training traces and refined via natural-language feedback improve multi-hop reasoning performance in long-context LMs across benchmarks and can be distilled into smaller models.
-
Scalable Multi Agent Diffusion Policies for Coverage Control
MADP uses diffusion models to generate interdependent actions for decentralized robot swarms in coverage control, trained via imitation from a clairvoyant expert and shown to generalize and outperform baselines across varying agent densities and importance densities.
-
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.
-
Foundation Models for Discovery and Exploration in Chemical Space
MIST models up to 10x larger than prior work, fine-tuned on over 400 structure-property tasks, match or exceed SOTA on benchmarks and demonstrate zero-shot olfactory perception mapping consistent with hyperbolic geometry.
-
Short window attention enables long-term memorization
Short sliding windows in hybrid attention-xLSTM models boost long-context performance by encouraging long-term memory use, and stochastic window sizing improves both short and long tasks.
-
Positional Encoding via Token-Aware Phase Attention
TAPA adds a learnable phase function to attention to preserve long-range token interactions, enabling direct continual pretraining, length extrapolation, lower perplexity, and stronger retrieval than RoPE-style methods.
-
F1: A Vision-Language-Action Model Bridging Understanding and Generation to Actions
F1 integrates next-scale visual foresight prediction into a Mixture-of-Transformer VLA architecture to reformulate action generation as foresight-guided inverse dynamics, achieving higher success rates on 136 tasks.
-
SpikingBrain: Spiking Brain-inspired Large Models
SpikingBrain-7B and SpikingBrain-76B achieve Transformer-comparable performance after continual pre-training on 150B tokens, with over 100x TTFT speedup on 4M-token sequences and 69.15% sparsity from event-driven spiking.
-
GLM-4.5V and GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning
GLM-4.5V reaches state-of-the-art results on 42 multimodal benchmarks among open-source models of similar size by applying reinforcement learning with curriculum sampling to a strong vision foundation model.
-
Learning Zero-Shot Subject-Driven Video Generation Using 1% Compute
A zero-shot subject-driven video generation framework that decomposes the task into identity injection from 200K subject-image pairs and motion preservation from 4K arbitrary videos, trained in 288 A100 GPU hours on CogVideoX-5B to match prior performance at 1% compute.
-
NVIDIA Nemotron 3: Efficient and Open Intelligence
NVIDIA releases the Nemotron 3 model family with hybrid Mamba-Transformer architecture, LatentMoE, NVFP4 training, MTP layers, and multi-environment RL post-training for reasoning and agentic tasks.
-
Video Parallel Scaling: Aggregating Diverse Frame Subsets for VideoLLMs
Video Parallel Scaling improves VideoLLM performance by aggregating outputs from parallel inferences on complementary disjoint frame subsets, effectively contracting the Chinchilla scaling law via uncorrelated visual evidence.
-
ShadowNPU: System and Algorithm Co-design for NPU-Centric On-Device LLM Inference
ShadowNPU presents shadowAttn, a co-designed sparse attention system that uses NPU pilot compute and techniques like graph bucketing and per-head sparsity to minimize CPU/GPU fallback during on-device LLM inference while maintaining accuracy.
-
Matrix-game 2.0: An open-source real-time and streaming interactive world model
Matrix-Game 2.0 introduces a scalable data pipeline, action-injection module, and few-step distillation to enable real-time streaming video generation at 25 FPS from game-engine interactions, with open-sourced weights and code.
-
Lightweight Domain Adaptation of a Large Language Model for Legal Assistance in the Indian Context
An 8B Llama model with RAG and prompt engineering scores 60.08% on the All-India Bar Examination, slightly above GPT-3.5 Turbo while claiming 22 times greater parameter efficiency via a new PEI metric.
-
Multi-Model Synthetic Training for Mission-Critical Small Language Models
Fine-tunes Qwen2.5-7B on 21,543 synthetic maritime Q&A pairs generated from 3.2B AIS records by GPT-4o and o3-mini, reaching 75% accuracy at 261x lower inference cost than larger models.
-
Phi-4-reasoning Technical Report
A 14B reasoning model trained via supervised fine-tuning on selected prompts and o3-mini traces, plus outcome RL, outperforms larger open models like DeepSeek-R1-Distill-Llama-70B on math, coding, planning and related benchmarks.
-
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.
-
Training LLMs on HPC Systems: Best Practices from the OpenGPT-X Project
Engineering report detailing HPC infrastructure, software choices, and performance measurements for training a 7B LLM using 3D parallelism on JUWELS Booster.
- Selective Rotary Position Embedding
- Robust Filter Attention: Self-Attention as Precision-Weighted State Estimation