Record · analyzed
AFlow: Automating Agentic Workflow Generation
Monte Carlo tree search over workflows written as code, using execution results to choose edits: 5.7% average gain over baselines.
1
Listed
Added to the list. How entries were checked when they were listed is in the journal.
2
Collected
| File | Source | SHA-256 |
|---|---|---|
| paper PDF, 2410.10762v42410.10762.pdf · 1.3 MB | arxiv.org/pdf/2410.10762v4 | 9be15f695f11dd5bc634c1c026bd2270eff3d3c4a53c4d9b51c012b7bd03d521 |
To check a file, download it from its source and compare its SHA-256 with the one above. Collected copies are kept in the editor's RSI system; manifest: design-life/literature/rsi-design/sources.json.
3
Analyzed
Quoted verbatim from design-life/research/rsi-design/evidence-review-2026-10-28.md (2026-10-28). Parts about applying the paper to the editor's private work are left out.
From “Evidence for effective RSI implementation” (2026-10-28)
Design choice: Evolve the representation
Supporting evidence: AFlow searches explicit agent structures; ACE maintains useful, incremental knowledge records. AFlow, ACE
Challenge or limit: AFlow also changes prompts and control code. ACE largely changes contents within a fixed structure. Neither isolates semantic task-schema evolution as the dominant source of improvement.
The evidence also leaves other productive changes available. DGM changes agent tools and software under fixed task criteria. AIDE² changes search and context policies. ACE changes reusable knowledge contents. AFlow changes prompts and connections. These interventions overlap, but they should not all be relabeled as improvements to the semantic task schema. DGM
The reviewed implementations often use a maximum search budget or convergence as operational stopping conditions; AFlow is an example. Those conventions do not validate satisfaction of Norman's quality condition. The implementation must preserve the distinction between an operational stop and an accepted result. It also needs a task-specific basis for nearly perfect, rather than relying on a proxy score alone. AFlow
4
Replicated
Not done yet. Not attempted yet.
5
Audited
Not done yet. Not attempted yet.
The editor's view
None recorded yet.