UltraEP is the first exact-load real-time expert balancer for large-EP MoE training and serving on rack-scale nodes, reaching 94.3% of ideal throughput and 1.49x over no-balancing.
super hub Canonical reference
Mixtral of Experts
Canonical reference. 80% of citing Pith papers cite this work as background.
319
Pith papers citing it
Background
80% of classified citations
abstract
We introduce Mixtral 8x7B, a Sparse Mixture of Experts (SMoE) language model. Mixtral has the same architecture as Mistral 7B, with the difference that each layer is composed of 8 feedforward blocks (i.e. experts). For every token, at each layer, a router network selects two experts to process the current state and combine their outputs. Even though each token only sees two experts, the selected experts can be different at each timestep. As a result, each token has access to 47B parameters, but only uses 13B active parameters during inference. Mixtral was trained with a context size of 32k tokens and it outperforms or matches Llama 2 70B and GPT-3.5 across all evaluated benchmarks. In particular, Mixtral vastly outperforms Llama 2 70B on mathematics, code generation, and multilingual benchmarks. We also provide a model fine-tuned to follow instructions, Mixtral 8x7B - Instruct, that surpasses GPT-3.5 Turbo, Claude-2.1, Gemini Pro, and Llama 2 70B - chat model on human benchmarks. Both the base and instruct models are released under the Apache 2.0 license.
hub tools
citation-role summary
background 50
baseline 4
method 3
dataset 2
other 2
citation-polarity summary
claims ledger
- abstract We introduce Mixtral 8x7B, a Sparse Mixture of Experts (SMoE) language model. Mixtral has the same architecture as Mistral 7B, with the difference that each layer is composed of 8 feedforward blocks (i.e. experts). For every token, at each layer, a router network selects two experts to process the current state and combine their outputs. Even though each token only sees two experts, the selected experts can be different at each timestep. As a result, each token has access to 47B parameters, but only uses 13B active parameters during inference. Mixtral was trained with a context size of 32k tok
authors
co-cited works
representative citing papers
HodgeCover isolates the harmonic kernel of a simplicial Laplacian on an expert 2-complex to identify irreducible merge cycles and selects experts for aggressive compression, matching or exceeding baselines on open-weight MoE models.
Sieve dynamically schedules MoE experts across GPU and PIM hardware to handle bimodal token distributions, achieving 1.3x to 1.6x gains in throughput and interactivity over static prior PIM systems on three large models.
ReLibra uses pre-known token-to-expert routing from RL rollouts to perform inter-batch expert reordering and intra-batch replication, delivering up to 1.6x higher throughput than Megatron-LM and 1.2x over oracle-equipped EPLB while staying within 6-10% of an ideal balanced baseline.
RoundPipe achieves near-zero-bubble pipeline parallelism for LLM training on consumer GPUs by dynamically dispatching computation stages round-robin, yielding 1.48-2.16x speedups and enabling 235B model fine-tuning on 8x RTX 4090.
Content-based routing succeeds only when models provide bidirectional context and perform pairwise comparisons, with bidirectional Mamba plus rank-1 projection reaching 99.7% precision at linear inference cost.
The AI Scientist framework enables LLMs to independently conduct the full scientific process from idea generation to paper writing and review, demonstrated across three ML subfields with papers costing under $15 each.
AgentClinic is a multimodal agent benchmark demonstrating that LLM diagnostic accuracy on MedQA drops to below one-tenth in sequential clinical simulations, with Claude-3.5 leading and large tool-use differences across models.
OSWorld provides the first unified real-computer benchmark for open-ended multimodal agent tasks, exposing large performance gaps between humans and state-of-the-art LLM/VLM agents.
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.
CrossPool separates weights and KV-cache into distinct GPU pools plus a planner, virtualizer, and layer-wise scheduler to cut P99 time-between-tokens by up to 10.4x versus prior kvcached multi-LLM systems.
For balanced Gaussian class projections, OOD AUROC is a linear function of MCS to the reference probe because both are sigmoid-shaped functions of the probe SNR on test data.
The paper constructs the VIBE benchmark and evaluates six visual in-context learning models on 14 datasets, 12 tasks, and 106 combinations under a unified one-shot protocol, revealing limitations and failure modes.
DICE formalizes multi-agent LLM coordination as discounted incomplete-information Markov games and introduces Heterogeneous Quantal Response Equilibrium (HQRE) to achieve unique stable equilibria with bounded regret, demonstrated via prompt-control and fine-tuning algorithms on eleven benchmarks.
MOSAIC is a simulation and DSE framework for heterogeneous NPUs that finds designs achieving 46.91% mean iso-area energy savings over homogeneous baselines on 20 workloads.
Introduces KINA benchmark with 899 items over 261 disciplines, formal (1-1/e) coverage guarantee and bonus-on-bar tournament theorem, plus evaluations of 42 models with top score 53.17%.
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.
ClinicalMC is a benchmark of 1,275 Chinese and 5,804 English multi-course clinical samples across four stages, evaluated via a multi-agent framework on closed-source, open-source, and medical LLMs in static and dynamic settings.
ViBE co-optimizes expert placement with measured GPU performance variability in MoE inference to cut execution-time imbalance, delivering 14% better SLO attainment and up to 45% lower P90 TTFT.
EST-PRM stress-tests five PRM models on 4,687 reasoning chains from MATH-500, GSM8K, and PRMBench using three label-preserving transformations and reports model-specific vulnerability patterns.
Introduces nexbax, a diagnostic framework with three themes and 10 dimensions for evaluating AI economic viability, operational practicality, and societal integrity in next-billion-user contexts.
Proposes the Intelligent Computing Architecture (ICA) as a six-layer framework with dual probabilistic-deterministic planes and three Amdahl-style heuristics to unify design of LLM-based systems.
Moral Trolley Arena shows frontier LLMs produce composite moral preferences that are compressed rather than additive functions of calibrated component act strengths across Moral Foundations Theory.
Introduces Latent Performance Profiling (LPP) as a task-agnostic framework deriving scalar metrics from LLM latent representations and dynamics to complement benchmark evaluations.
citing papers explorer
-
RULER: What's the Real Context Size of Your Long-Context Language Models?
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.
-
Every Act Has Its Price: Compressed Moral Composition in Frontier LLMs
Moral Trolley Arena shows frontier LLMs produce composite moral preferences that are compressed rather than additive functions of calibrated component act strengths across Moral Foundations Theory.
-
Latent Performance Profiling of Large Language Models
Introduces Latent Performance Profiling (LPP) as a task-agnostic framework deriving scalar metrics from LLM latent representations and dynamics to complement benchmark evaluations.
-
Large Language Model Selection with Limited Annotations
SELECT-LLM is the first active model selection framework for LLMs that uses expected information gain from pairwise output similarities to minimize required annotations, reporting up to 84.78% cost reduction across 23 datasets and 156 models.
-
Fine-grained Claim-level RAG Benchmark for Law
ClaimRAG-LAW is a French-English legal RAG benchmark with claim-level granularity for experts and non-experts that reveals limitations in current retrieval and generation performance.
-
HEBATRON: A Hebrew-Specialized Open-Weight Mixture-of-Experts Language Model
Hebatron is the first open-weight Hebrew MoE LLM adapted from Nemotron-3, reaching 73.8% on Hebrew reasoning benchmarks while activating only 3B parameters per pass and supporting 65k-token context.
-
GTA-2: Benchmarking General Tool Agents from Atomic Tool-Use to Open-Ended Workflows
GTA-2 benchmark shows frontier models achieve below 50% on atomic tool tasks and only 14.39% success on realistic long-horizon workflows, with execution harnesses like Manus providing substantial gains.
-
TEAM: Temporal-Spatial Consistency Guided Expert Activation for MoE Diffusion Language Model Acceleration
TEAM accelerates MoE dLLMs up to 2.2x by exploiting temporal-spatial consistency in expert routing to accept more tokens with fewer activations.
-
ModeX: Evaluator-Free Best-of-N Selection for Open-Ended Generation
ModeX selects the modal semantic output from multiple LLM generations via a similarity graph and recursive spectral clustering without needing reward models or evaluators.
-
Soft Head Selection for Injecting ICL-Derived Task Embeddings
SITE applies soft gradient-based head selection to inject ICL-derived task embeddings, outperforming prior embedding adaptation and few-shot ICL across generation, reasoning, and NLU tasks on 12 LLMs from 4B to 70B parameters.
-
Jamba: A Hybrid Transformer-Mamba Language Model
Jamba presents a hybrid Transformer-Mamba MoE architecture for LLMs that delivers state-of-the-art benchmark performance and strong results up to 256K token contexts while fitting in one 80GB GPU with high throughput.
-
Massive Activations in Large Language Models
Massive activations are constant large values in LLMs that function as indispensable bias terms and concentrate attention probabilities on specific tokens.
-
OLMo: Accelerating the Science of Language Models
OLMo delivers a fully open competitive language model with training data, code, and evaluations to enable community-driven scientific research on LMs.
-
MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries
MultiHop-RAG is a new benchmark dataset demonstrating that existing retrieval-augmented generation systems perform poorly on multi-hop queries requiring retrieval and reasoning over multiple evidence pieces.
-
CLExEval: A Human-in-the-Loop Framework for Qualitative Evaluation of LLM Clinical Reasoning
CLExEval introduces a human-annotated evaluation framework on 40 rare cases that identifies verbosity bias, hidden knowledge paradox, and 68.6% reasoning-to-output mismatch in LLMs while showing LLM-as-a-Judge overestimates reliability.
-
Wait, am I Being Fair? Characterizing Deductive Stereotyping and Mitigating It with Fair-GCG
The paper characterizes deductive stereotyping in LLMs and introduces Fair-GCG to discover injection phrases that improve fairness across benchmarks, reasoning, and real-world tasks.
-
Customer-Agent: Overcoming Context Limitations in Ultra-Long Shopping Trajectories via Tool-Augmented Agents and RLVR
Introduces ShopTrajQA long-context benchmark and an RLVR-trained tool-augmented agent that bypasses LLM context limits by external file storage and code-based retrieval for shopping trajectories.
-
When New Generators Arrive: Lifelong Machine-Generated Text Attribution via Ridge Feature Transfer
RidgeFT enables replay-free lifelong MGT attribution via frozen encoder, class-wise sufficient statistics, covariance calibration, and closed-form ridge regression updates, outperforming baselines on macro-F1 and retention-adaptation balance.
-
DLLG: Dynamic Logit-Level Gating of LLM Experts
DLLG learns token-level fusion weights for LLM experts from sparse response supervision and outperforms routing, ensembling, and merging baselines on reasoning and code tasks.
-
Expert-Aware Refusal Steering
Refusal steering works on MoE LLMs; expert-aware variants succeed with single-expert outputs and refusal signals differ from routing patterns.
-
On the Limits of LLM Adaptability: Impact of Model-Internalized Priors on Annotation Task Performance
LLMs correct only 34.8% of zero-shot annotation errors via prompting, and Definition-Specific Familiarity correlates positively with performance (partial r = +0.41) while memorization metrics do not.
-
ReverseMath: Answer Inversion for Scalable and Verifiable Mathematical Problem Generation
ReverseMath uses answer inversion to generate paired original and reversed math problems with known answers for detecting memorization and improving LLM reasoning via data augmentation.
-
Artificial Intolerance: Stigmatizing Language in Clinical Documentation Skews Large Language Model Decision-Making
Frontier LLMs exhibit bias from stigmatizing language in clinical vignettes across four conditions, skewing decisions toward less aggressive management, with limited mitigation from Chain-of-Thought or self-debiasing prompts.
-
HyDRA: Hybrid Dynamic Routing Architecture for Heterogeneous LLM Pools
HyDRA routes queries to cost-effective LLMs by predicting multi-dimensional capability requirements with a multi-head encoder and applying shortfall matching against configuration-defined model profiles, delivering up to 72.5 percent cost savings on coding benchmarks while remaining decoupled from具体
-
From Text to Voice: A Reproducible and Verifiable Framework for Evaluating Tool Calling LLM Agents
A dataset-agnostic framework converts text tool-calling benchmarks to paired audio evaluations via TTS, speaker variation and noise, then evaluates seven omni-modal models showing model- and task-dependent performance with small text-to-voice gaps.
-
Instructions Shape Production of Language, not Processing
Instructions trigger a production-centered mechanism in language models, with task-specific information stable in input tokens but varying strongly in output tokens and correlating with behavior.
-
LegalCiteBench: Evaluating Citation Reliability in Legal Language Models
LegalCiteBench reveals that current LLMs achieve under 7% accuracy on closed-book legal citation retrieval and completion tasks, with misleading answer rates above 94% for nearly all tested models.
-
PaT: Planning-after-Trial for Efficient Test-Time Code Generation
PaT defers planning until after failed trials in LLM code generation, enabling heterogeneous cheap-plus-powerful model setups that match large-model performance at roughly 69% lower cost.
-
EMO: Pretraining Mixture of Experts for Emergent Modularity
EMO pretrains MoEs using document boundaries to induce semantic expert specialization, enabling modular subset deployment with minimal accuracy loss unlike standard MoEs.
-
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.
-
AGoQ: Activation and Gradient Quantization for Memory-Efficient Distributed Training of LLMs
AGoQ delivers up to 52% lower memory use and 1.34x faster training for 8B-32B LLaMA models by using near-4-bit adaptive activations and 8-bit gradients while preserving pretraining convergence and downstream accuracy.
-
Test-Time Safety Alignment
Optimizing input embeddings sub-lexically via black-box zeroth-order gradients neutralizes all safety-flagged responses from aligned models on standard benchmarks.
-
Culture-Aware Machine Translation in Large Language Models: Benchmarking and Investigation
LLMs exhibit large performance gaps in culture-aware translation, translation strategies systematically affect outputs, culture-specific items vary in difficulty, and models recognize cultural knowledge better than they use it correctly in translations.
-
Beyond N-gram: Data-Aware X-GRAM Extraction for Efficient Embedding Parameter Scaling
X-GRAM applies data-aware dynamic token injection with hybrid hashing and local feature extraction to achieve up to 4.4 accuracy point gains over vanilla backbones and 3.2 over retrieval baselines at 0.73B-1.15B scales using 50% smaller tables.
-
Towards Robust Real-World Spreadsheet Understanding with Multi-Agent Multi-Format Reasoning
SpreadsheetAgent uses incremental multi-format reading, structural sketching, and verification to raise spreadsheet benchmark accuracy from 35.27% to 38.16%.
-
Cram Less to Fit More: Training Data Pruning Improves Memorization of Facts
Loss-based pruning of training data to limit facts and flatten their frequency distribution enables a 110M-parameter GPT-2 model to memorize 1.3 times more entity facts than standard training, matching a 1.3B-parameter model on the full dataset.
-
Why Attend to Everything? Focus is the Key
Focus learns a few centroids to gate long-range token attention, producing sparse attention that matches or beats full attention quality with up to 8.6x speedup at million-token lengths.
-
ShinkaEvolve: Towards Open-Ended And Sample-Efficient Program Evolution
ShinkaEvolve improves sample efficiency in LLM-driven program evolution via parent sampling, code novelty rejection-sampling, and bandit LLM ensemble selection, achieving new SOTA circle packing with 150 samples and gains on math reasoning and competitive programming tasks.
-
Training-Free Multimodal Large Language Model Orchestration
LLM Orchestration integrates modality experts via an LLM controller, cross-modal memory, and interaction layer to enable multimodal input-output without gradient-based training.
-
Mixture-of-Experts Can Surpass Dense LLMs Under Strictly Equal Resource
MoE models with activation rates in an optimal region outperform dense LLMs of identical total parameter count, training compute, and data budget, with the optimal region consistent across scales.
-
Mixture-of-Transformers: A Sparse and Scalable Architecture for Multi-Modal Foundation Models
MoT decouples non-embedding parameters by modality in transformers to match dense multi-modal performance with roughly one-third to one-half the FLOPs.
-
Dictionary Insertion Prompting for Multilingual Reasoning on Multilingual Large Language Models
DIP interleaves English word translations into non-English prompts to boost multilingual reasoning on synthetic benchmarks spanning 10-200 languages.
-
LaMSUM: Amplifying Voices Against Harassment through LLM Guided Extractive Summarization of User Incident Reports
LaMSUM is a novel multi-level LLM framework with voting methods for extractive summarization of large incident report collections that outperforms prior extractive methods.
-
Mixture-of-Agents Enhances Large Language Model Capabilities
A layered Mixture-of-Agents system combining multiple LLMs achieves state-of-the-art results on AlpacaEval 2.0 (65.1%), MT-Bench, and FLASK, outperforming GPT-4 Omni.
-
MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark
MMLU-Pro is a revised benchmark that makes language model evaluation harder and more stable by using ten options per question and emphasizing reasoning over simple knowledge recall.
-
Chameleon: Mixed-Modal Early-Fusion Foundation Models
Chameleon is an early-fusion token model that handles mixed image-text sequences for understanding and generation, achieving competitive or superior performance to larger models like Llama-2, Mixtral, and Gemini-Pro on captioning, VQA, text, and image tasks.
-
Argumentative Large Language Models for Explainable and Contestable Claim Verification
ArgLLMs build argumentation frameworks from LLMs to support explainable and contestable formal reasoning for claim verification.
-
It Takes One to Bias Them All: Breaking Bad with One-Shot GRPO
One-shot GRPO on a single biased example induces generalizing stereotype bias in post-trained LLMs, with susceptibility varying by initial bias likelihood.
-
Parameter Alignment Mitigates Catastrophic Forgetting in Multilingual Expert Language Models
Parameter alignment strategies substantially reduce forgetting in family-based continual pretraining of multilingual LLMs across 32 languages with minimal impact on language acquisition.
-
TIDE: Efficient and Lossless MoE Diffusion LLM Inference with I/O-aware Expert Offload
TIDE schedules I/O-aware expert offloading for MoE diffusion LLMs by solving for an optimal refresh interval that exploits temporal stability of activations, yielding up to 1.5x throughput gain losslessly.