English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Recursive Agent Harnesses: Spawning Full Sub-Agents Pushes Long-Context Reasoning from 71% to 89%

Forum topic · ✨步子哥 · 2026-06-13

Summary

Recursive Agent Harnesses (RAH) is a new agent architecture in which a parent agent recursively spawns complete agent harnesses—each with its own tools, file system, and reasoning capability—to handle subtasks. Unlike recursive language models (RLMs), whose recursion unit is a bare model call without tool access, and coding agents, which rely on regex-based extraction, RAH's recursion unit is a full harness. On the Oolong-Synthetic long-context benchmark (199 samples, 1K to 4M tokens), RAH with GPT-5 as the backbone scores 81.36% accuracy versus 71.75% for a Codex-style coding agent and 64.38% for RLM; with Claude Sonnet 4.5, it reaches 89.77%. RAH supports code-execution spawning for large batches of items and JSON tool-call spawning for small ones. Key findings: gains stem from the harness design rather than the model, improvements hold across all context lengths, and the choice of recursion unit matters as much as model choice. Limitations include higher compute cost, latency, failure on inter-item dependencies, and validation only on a structured benchmark. Paper by Elias Lumer et al., PwC US (arXiv:2606.13643).

How do you read a 536K-token document?

Imagine a 536K-token document containing 1,772 key-value pairs. Your task: extract every entry and answer questions about them.

  • A coding agent writes a regex script and loops through entries. Fast—but regex only captures surface patterns and fails on entries that require reasoning.
  • A recursive language model (RLM) chunks the document and reasons over each piece. It can reason—but it can't see a file system or use tools.
  • What if you spawned a clone for each entry, each with a full toolbox, file system, and reasoning ability? That is the core idea of the Recursive Agent Harness (RAH).
  • What is a "recursive agent harness"?

    Breaking down the name:

  • Harness: everything that turns an LLM into a working agent—tools, file system, context engineering, orchestration logic. Like a car wiring harness connecting the engine, steering, and dashboard into a drivable vehicle.
  • Recursive: the harness can produce harnesses. A parent agent writes code that launches a complete sub-agent harness for each subtask.
  • Key difference: RLM's recursion unit is a *model call* (no tools); RAH's recursion unit is a *full harness* (tools, file system, reasoning).
  • Analogy: RLM is one person breaking a big task into pieces and doing them alone; RAH is a manager delegating to a team of interns, each with a full desk and toolbox.

    Two spawning modes

  • Code-Execution Spawning: for large numbers of items, the parent writes an executable script that launches a sub-agent per item via a Task() function, running them in parallel—like a factory line with one dispatcher writing instructions.
  • JSON Tool-Call Spawning: for only 1–5 items, launch sub-agents directly through structured function calls—like a boss verbally assigning a few tasks.
Sub-agents inherit the spawning ability and can recurse further, up to a depth limit.

Results: the harness wins

Tested on Oolong-Synthetic (199 samples, 13 context-length buckets from 1K to 4M tokens), with GPT-5 fixed as backbone for fair comparison:

| Method | Accuracy | |--------|----------| | Codex (coding agent, regex) | 71.75% | | RLM (recursive LM, no tools) | 64.38% | | RAH (recursive agent harness) | 81.36% |

With Claude Sonnet 4.5 as backbone, RAH reaches 89.77%.

Key findings:

1. Gains come from the harness, not the model: same GPT-5, from 71.75% to 81.36%, purely via architecture. 2. Consistent improvement across all length buckets, including 4M-token documents. 3. Coding agents bottleneck on reasoning: regex is fast but not smart. 4. RLMs bottleneck on tools: they can reason but can't see files—like doing math blindfolded.

Why it matters

RAH exposes an overlooked design dimension: the choice of the recursion unit. Upgrading it from a bare model call to a full harness gives you both reasoning and tool access—the half that coding agents and RLMs each lack. Moreover, RAH's spawning logic is ordinary program code, not a fixed recursion convention or predefined tool schema, letting the parent parameterize concurrency, per-item instructions, and output paths in the same language it writes code in. Anthropic's dynamic workflows already use a similar code-driven spawning pattern in production; RAH is the first systematic naming, definition, and evaluation of it.

Honest assessment

RAH is not a silver bullet. It costs more compute and latency—launching a full sub-agent per item is not cheap. The authors acknowledge failure modes: when items have dependencies, parallel spawning can miss cross-item reasoning chains. RAH has only been validated on Oolong-Synthetic, a highly structured (key-value) benchmark; performance on more open-ended long-context tasks remains untested.

Still, the core insight is clear: in agent architectures, the choice of recursion unit matters as much as the choice of model. The same backbone, with a different recursion strategy, gains nearly 10 percentage points—a signal worth attention for anyone building agent systems.

---

Paper: Recursive Agent Harnesses Authors: Elias Lumer, Sahil Sen, Kevin Paul, Vamse Kumar Subbiah (PwC US) Field: Multi-Agent Systems, Long-Context Reasoning

Tags

#ai-agents#long-context#recursive-agents#multi-agent-systems#llm-reasoning#agent-architecture#gpt-5#paper-review

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/177981264