Blog page on multi-agent architecture has no article text
The fetched page shows only a contact footer, not the architecture patterns named in the title.
Stories per day · 30 days
1 RSS
22 stories, newest first · one card per story
Stories tagged #architecture
Huginn-style looped models beat Ouro at matched scale
A Huginn-style MoE uses 75% fewer resident parameters than a 112-layer feedforward baseline under matched FLOPs.
Copy-Paste Language Models
Copy-Paste Language Models add a lightweight copy head that improves copying, efficiency, and long-context performance.
This AI startup says it can generate permit-ready blueprints in minutes
Startup Blueprints AI says it generates permit-ready architectural drawings from sketches in minutes.
2
Multi-Stream Attention Residuals
MSAR adds multi-head depth attention to residuals, with a 1.58× iso-loss speedup over standard residuals.
Recurrent Looped Transformer
Recurrent Looped Transformers track algorithmic state far beyond training length via per-token feedback.
Ensure AutoCheck prepend is last in module templates and update archi…
Metasploit templates now place AutoCheck prepend last and update architecture handling.
Do Language Models Need a Trainable Input Embedding Table? Fixed Minimal Token Codes at 1.7B-Class Scale
Fixed 16-bit token codes let 1.7B decoder-only models learn substantial language ability without trainable input embeddings.
Beyond the Parameter Monolith: Reconstructive Memories, Executable Skills, and Residual Assembly for Language Models
FEM-ASM separates model storage, executable skills, and neural coordination, with limited experimental results.
Looping Beyond Twice: A Scalable Recipe for Looped Mixture-of-Experts
LOOM scales looped Mixture-of-Experts Transformers to 9-12 loops, cutting 1.7B model perplexity from 9.62 to 7.77 with accuracy gains.
Triadic Linear Attention: Three-Dimensional Recurrent States for Long-Context Sequence Modeling
Triadic linear attention extends linear attention to third-order tensor states, improving long-context modeling and recall over state-enlargement alternatives.
What Makes Recurrence Effective in Looped Language Models?
Looped language models gain reasoning from recurrence beyond training, but extra loops can hurt knowledge.
How to Loop MoE: Flatten the Experts, Untie the Attention
Foil flattens experts and unties attention in looped MoE models, lowering pretraining loss at equal parameters and compute.
FactorEngram: Factorized N-gram Memory with Basis-Level Gating for Language Models
FactorEngram adds factorized n-gram memory with basis-level gating to language models.
WhiteMatter: All-to-All Cross-Layer Connections via KV Source Mixing
WhiteMatter mixes key-value sources across Transformer layers, matching deeper models with a smaller KV cache.
Change the Product, Keep the Parameters: Associative Algebra Layers for Transformers
Algebraic Transformer layers raise small-model generation throughput 6.2-7.8% but lower downstream scores.
IntBMoE: Integrating Block-Level Conditioning into Expert Composition for Full-Participation Mixture-of-Experts
IntBMoE is a block-conditioned Mixture-of-Experts architecture achieving full participation with sparse execution, deployed in AMap's recommendation system.
Agentic Societies Need a Social Harness
Researchers propose a layered 'social harness' to stop malicious AI agents from exploiting inter-agent communication in multi-agent societies.
Authorization Architectures for Tool-Using AI Agents
Review paper proposes an authorization reference architecture for tool-using AI agents, identifying runtime enforcement and delegation bounds as unresolved gaps.
Don't Trust the Super-App: A Case Study of Russia's Max
Researchers show super-apps like Russia's Max can silently capture mini-app UI, storage, network traffic, and authentication, enabling untraceable user impersonation.
Kalman Delta Networks: Uncertainty-aware Associative Memory
Kalman Delta Networks add uncertainty tracking to linear-attention associative memory, improving perplexity and downstream accuracy at 750M and 1.3B scales.
Graph Machine: Towards Better Pretraining via Edges
Researchers propose Graph Machine, an O(n)-state sparse architecture that replaces 75% of Qwen3-0.6B dense layers with only slight loss change.