SYNTHESIS NOTE
Topics›Looped Models›this note

Can reasoning systems scale faster by exploring parallel paths instead?

Current recursive reasoning models refine a single latent trajectory deeply, which is slow. Could sampling multiple trajectories in parallel achieve better reasoning with lower latency, and would that scale differently than serial refinement?

Synthesis note · 2026-05-28 · sourced from Looped Models

Recursive Reasoning Models (RRMs) increase reasoning capability by iterating a shared transition function over a latent state — more iterations means more "thinking" without extending the output sequence. This is depth scaling, and it decouples reasoning depth from both parameter count and output length. GRAM (Generative Recursive reAsoning Models) argues this is only half the story: depth alone is insufficient because a single refinement path can become trapped in a suboptimal trajectory, and many problems have ambiguity or multiple valid solutions that a single converging path cannot represent.

The structural claim is that future recursive reasoners should be not only deep (repeated refinement) but also wide (maintaining and exploring multiple latent trajectories in parallel). GRAM operationalizes width by turning the latent transition stochastic and sampling several trajectories simultaneously. Crucially, width sidesteps the latency penalty that depth-only scaling incurs: sampling N trajectories runs in parallel, whereas adding N refinement steps is serial and accumulates wall-clock time.

This reframes the inference-scaling design space for latent architectures. It mirrors at the latent-state level what parallel-vs-sequential debates established at the token level — since Why does parallel reasoning outperform single chain thinking?, breadth often beats depth under a fixed budget because independent paths sample the solution distribution rather than inflating variance along one path. GRAM brings that lesson inside the recurrent block, where prior work like Can models reason without generating visible thinking tokens? had only scaled depth. The counterpoint to watch: since Can parallel architectures solve inherently sequential problems?, width cannot substitute for depth on inherently serial problems — the two axes are complements, not interchangeable knobs. Why it matters: it gives latent reasoning a second, latency-cheap scaling dimension and explains why deterministic RRMs underperform on multi-solution tasks.

Inquiring lines that read this note 142

This note is a source for these research framings, grouped by the broader line of inquiry each explores. Scan the bold lines of inquiry; follow any specific question forward.

What determines whether visible reasoning traces improve task performance? When does routing outperform scaling as a performance solution? How can recommendation systems balance individual and population-level goals? How does model scale affect fine-tuning and downstream task performance? Do base models contain latent reasoning capabilities that training can activate? Does reinforcement learning create new reasoning capabilities or refine existing ones? When do multi-agent systems outperform single agents? How does scaling test-time compute compare to increasing model size? When do additional thinking tokens stop improving reasoning performance? Does parallel reasoning outperform sequential approaches in reasoning efficiency? Do reasoning traces accurately represent what models actually compute? Does evolutionary optimization at test time improve model performance effectively? Does architectural recurrence enable computational capabilities that fixed-depth models cannot achieve? How can parameter sparsity improve interpretability while maintaining model capability? How do workflow position and structure enable attacks in multi-agent systems? How should test-time compute be allocated adaptively across task difficulties? How does policy entropy collapse limit reasoning RL scaling? Why does self-revision increase model confidence despite degraded accuracy? How does silent agreement emerge and undermine deliberation in multi-agent systems? How do pretraining data composition and task diversity shape reasoning? Does training data format shape reasoning more than domain content? What makes reasoning models struggle with ill-posed and unfamiliar problems? Does reasoning architecture matter more than model scale? Does reasoning fine-tuning trade off other critical model abilities? Can scaling alone produce compositional generalization in neural networks? Does chain-of-thought reasoning faithfully reflect actual model reasoning processes? Why does separating planning from execution enhance reasoning performance? What actually drives chain-of-thought performance gains in reasoning tasks? Should reward models emphasize process understanding or only final outcomes? How can language models effectively manage long-context memory without quality degradation? Does semantic diversity improve reasoning quality in reinforcement learning? How do different post-training structures affect learning outcomes?

Related concepts in this collection 4

This note in its neighbourhood — explore the map, then jump to a related concept in the list below.

Concept map
13 direct connections · 115 in 2-hop network ·medium cluster Open in graph ↗
Can reasoning systems scale faster by exploring … Why does parallel reasoning outperform single chai… Can models reason without generating visible think… Can parallel architectures solve inherently sequen… Can we explore multiple reasoning paths without co…

Click a node to walk · click center to open · click Open in graph to see this note in the full knowledge graph

your link semantically near linked from elsewhere

Related papers in this collection 8

Papers most semantically related to this note, ranked by cosine similarity in the embedding space.

Search by related questions 4

Suggested questions this note speaks to — click to search the collection, or type your own.

Why does parallel thinking outperform sequential thinking under the same token budget?Which problems cannot be solved by parallel architectures and require serial depth?How do continuous concept tokens explore multiple reasoning paths without explicit sampling?Can latent reasoning scale test-time compute without verbalized tokens or special training?
Original note title

reasoning systems should scale in width by sampling parallel latent trajectories not only in depth