A bipartite factor graph with message-passing protocol and asymmetric damping aggregates multi-LLM predictions, cutting token use by 97% and API calls by 6X while outperforming baselines on MMLU, MMLU-Pro, GPQA, and MedMCQA.
hub Canonical reference
arXiv preprint arXiv:2511.20639 , year=
Canonical reference. 83% of citing Pith papers cite this work as background.
17
Pith papers citing it
Background
83% of classified citations
abstract
Multi-agent systems (MAS) extend large language models (LLMs) from independent single-model reasoning to coordinative system-level intelligence. While existing LLM agents depend on text-based mediation for reasoning and communication, we take a step forward by enabling models to collaborate directly within the continuous latent space. We introduce LatentMAS, an end-to-end training-free framework that enables pure latent collaboration among LLM agents. In LatentMAS, each agent first performs auto-regressive latent thoughts generation through last-layer hidden embeddings instead of text. Then, a shared latent working memory preserves and transfers each agent's internal representations and latent thoughts, ensuring lossless information exchange without re-encoding. We provide detailed theoretical analyses showing that LatentMAS achieves higher expressiveness and lossless information preservation with lower overall complexity than standard text-based MAS. In addition, empirical evaluations across 9 comprehensive benchmarks spanning math and science reasoning, commonsense understanding, and code generation show that LatentMAS outperforms advanced single agents and text-based MAS baselines, achieving up to 14.6% higher accuracy, reducing output token usage by 70.8%-83.7%, and providing 4$\times$-4.3$\times$ faster end-to-end inference. Code and data are fully open-sourced at https://github.com/Gen-Verse/LatentMAS.
hub tools
citation-role summary
background 5
extension 1
citation-polarity summary
years
2026 17representative citing papers
A survey that unifies prior work on multi-agent LLM systems via the LIFE framework, mapping dependencies across collaboration, failure attribution, and autonomous self-evolution while identifying cross-stage challenges.
TFlow enables multi-agent LLMs to collaborate via transient low-rank LoRA perturbations derived from sender activations, yielding up to 8.5 accuracy gains and 83% token reduction versus text-based baselines on Qwen3-4B models.
LaTER reduces LLM token usage 16-33% on reasoning benchmarks by exploring in latent space then switching to explicit CoT verification, with gains like 70% to 73.3% on AIME 2025 in the training-free version.
LACO introduces Iterative Latent Deliberation, Cross-Horizon Saliency Attribution, and Structured Semantic Knowledge Distillation to enable low-latency latent communication in collaborative driving while preserving performance in CARLA simulations.
VerifyMAS improves failure attribution in LLM multi-agent systems via hypothesis verification on full trajectories, error taxonomy-based data construction, and fine-tuned verifier models, outperforming prior direct-prediction methods on Aegis-Bench and Who&When.
KV-Fold turns frozen transformers into stable long-context models by folding the KV cache across sequence chunks in repeated forward passes.
MASPO jointly optimizes prompts in multi-agent LLM systems via downstream-success evaluation and evolutionary beam search, delivering 2.9 average accuracy gains over prior methods across six tasks.
QKVShare enables efficient quantized KV-cache handoff for on-device multi-agent LLMs, cutting TTFT versus re-prefill across tested contexts while adaptive quantization stays competitive with uniform baselines on GSM8K.
A large model generates a compact reasoning signal that a small model uses to solve tasks, reducing the large model's output tokens by up to 60% on benchmarks like AIME and GPQA.
MeloTune implements learned per-listener Personal Arousal Functions and mesh memory protocols on mobile devices to predict affective trajectories and enable peer-coupled proactive music selection, reporting 96.6% pattern accuracy in deployment.
LLM agent committees exhibit representational collapse with mean cosine similarity of 0.888, and diversity-aware consensus reaches 87% accuracy on GSM8K versus 84% for self-consistency at lower cost.
Zeroth-order optimization is underexplored rather than underpowered in deep learning, with limitations stemming from full-space designs that can be addressed via subspace, spectral, and systems-aware approaches.
Eywa enables language-based agentic AI systems to collaborate with specialized scientific foundation models for improved performance on structured data tasks.
The paper delivers a unified survey of token economics for LLM agents, conceptualizing tokens as production factors, exchange mediums, and units of account across micro, meso, macro, and security dimensions using established economic theories.
This survey organizes RL for LLM multi-agent systems into reward families, credit units, and five orchestration sub-decisions, notes the absence of explicit stopping-decision training in its paper pool, and releases a tagged corpus.
The paper analyzes evolving security and safety threats in generative AI from content generation to agentic actions, noting that attack surfaces expand faster than defenses and that many safeguards require institutional coordination not yet in place.
citing papers explorer
-
KV-Fold: One-Step KV-Cache Recurrence for Long-Context Inference
KV-Fold turns frozen transformers into stable long-context models by folding the KV cache across sequence chunks in repeated forward passes.
-
When Less is Enough: Efficient Inference via Collaborative Reasoning
A large model generates a compact reasoning signal that a small model uses to solve tasks, reducing the large model's output tokens by up to 60% on benchmarks like AIME and GPQA.
-
Representational Collapse in Multi-Agent LLM Committees: Measurement and Diversity-Aware Consensus
LLM agent committees exhibit representational collapse with mean cosine similarity of 0.888, and diversity-aware consensus reaches 87% accuracy on GSM8K versus 84% for self-consistency at lower cost.
-
Position: Zeroth-Order Optimization in Deep Learning Is Underexplored, Not Underpowered
Zeroth-order optimization is underexplored rather than underpowered in deep learning, with limitations stemming from full-space designs that can be addressed via subspace, spectral, and systems-aware approaches.