Back to Ruflo

Performance SOTA Report — 2026-07-05

v3/docs/research/dream-cycle-2026-07-05-performance.md

3.34.05.9 KB
Original Source

Performance SOTA Report — 2026-07-05

TL;DR: PolyKV (arXiv:2604.24971, Grade A) proves a shared asymmetrically-compressed KV pool slashes multi-agent memory by 97.7% at 15 concurrent agents — Ruflo has no cross-agent shared KV pool and this is the single highest-leverage performance gap for 2026.


What's New in 2026

FindingSourceConfidence
PolyKV: shared KV pool (int8 keys + FWHT+3-bit Lloyd-Max values) cuts 15-agent memory 19.8 GB → 0.45 GB (97.7%), 2.91x compression, BERTScore F1 0.928arXiv:2604.24971 (Patel & Joshi, Apr 2026)A
ConServe disaggregated prefill/decode scheduling reduces p95 TTFT by 51.08% and energy by 7.51%arXiv Jun 2026 (Ding, Hosseini et al.)A
AsymCache multi-segment KV eviction cuts TTFT 1.90–2.03x and TPOT 1.62–1.71xarXiv Jun 2026 (Shi, Chen et al.)A
IBM speculative decoding + FP8: 4.48x throughput speedup with maintained task performancearXiv:2606.18502 (Dashore et al., IBM, Jun 2026)B
CLSR symbolic inter-agent language: 3–6x fewer tokens in agent-to-agent messages vs chain-of-thoughtarXiv Jun 2026 (Pei, Huang, Wang)B

Ruflo Current Capability

CapabilityCurrent StateGap
KV cachePer-agent cache, referenced in CLAUDE.mdNo cross-agent shared/compressed pool
Speculative decodingNot mentioned in architectureAbsent; 4.48x throughput unrealized
Disaggregated schedulingNot addressedAbsent; 51% p95 TTFT reduction unrealized
Inter-agent communicationSendMessage text protocolNo symbolic/compressed format; CLSR 3–6x savings unrealized
Vector searchHNSW (1.9x–4.7x measured vs brute force)Addressed; not performance gap

Competitor Comparison

FrameworkKey 2026 Performance ClaimSourceGrade
LangGraph10,155ms avg latency, 2.70 rps, 30–40% fewer tokens than CrewAI on medium tasksdev.to benchmark 2026; AI Dev Day IndiaB
AutoGen (AG2)~56,700 tokens / 4-agent 5-round debate; 4–5x overhead vs LangChain on complex tasks; GA Feb 2026f3fundit.com; secondtalent.com 2026C
CrewAI43/100 token efficiency score; ~32s/run; 44% task failure rate in Rust framework benchmarkdev.to audit; aimultiple.com 2026C
OpenAI Agents SDK"Fastest latency and highest token efficiency" in one enterprise benchmark; Swarm deprecated Mar 2026qubittool.com; selecthub.com 2026B
Ruflo (claude-flow)HNSW 1.9x–4.7x measured; SONA 0.0043ms/adaptCLAUDE.md / benchmark-intelligence.mjsA

Benchmarks

BenchmarkMetricValueGrade
PolyKV — Llama-3-8B, 15 agents, 4K ctxKV memory reduction19.8 GB → 0.45 GB (97.7%)A
PolyKVCompression ratio2.91x stableA
PolyKVQuality (BERTScore F1)0.928 (perplexity +0.57%)A
ConServep95 TTFT reduction51.08%A
ConServeEnergy efficiency gain7.51%A
AsymCacheTTFT improvement1.90–2.03xA
AsymCacheTPOT improvement1.62–1.71xA
IBM Spec+FP8Throughput speedup4.48xB
CLSR symbolic commsToken count reduction3–6xB
LangGraphTask completion (Qwen3 32B, M4 Max)62% vs AutoGen 58%, CrewAI 54%C (community benchmark)

Scan Findings

Security (Grade A)

MESA (arXiv:2606.30602, Li et al., Jun 2026): a single compromised agent-to-agent communication edge accounts for up to 75% of total attack success in multi-agent systems. AgentFlow (arXiv:2607.01640, Wang et al., Jul 2026) introduces static taint analysis via Agent Dependency Graphs covering MCP server pipelines. AI-Infra-Guard provides 75+ AI component rules and LLM-driven MCP server auditing. The 2025 OWASP LLM Top 10 "Excessive Agency" threat is now operationalized through supply-chain and memory-poisoning benchmarks. Ruflo gap: no static taint analysis or agent-to-agent channel integrity verification.

Hive-Mind / Byzantine Consensus (Grade B)

arXiv:2602.02170 (Rodriguez & Díaz, Feb 2026) evaluates six BFT protocols under f < n/3 and confirms bounded self-modification of coordination protocols is formally auditable. Broader collective-intelligence work for LLM hive-mind systems has not yet appeared at scale — the field is mapping BFT theory to practice. Ruflo's existing byzantine-coordinator and raft-manager are ahead of most published work.


SOTA Proof & Witness

Session commit: a5f86ad0ada8aca3e8f664202a452714355990f5
Report SHA-256: 39545a8bb7283d10e370d3418584165101762c8550982e6cd069868c23181383
Witness stamp: 8c1959eec4d5b534b3d3c6030617c2bfe2c109520903ed435d6bdfdaa34ea90b
Verification: cat dream-gist-2026-07-05.md | sha256sum → concat with session commit → sha256sum → must equal witness stamp.


  1. ADR-182: Cross-Agent Shared KV Pool (PolyKV architecture) — Implement a shared asymmetrically-compressed KV pool using int8 key quantization and FWHT+Lloyd-Max value compression. Target: reduce 15-agent KV memory by ≥90% (from ~20 GB to <2 GB). Implementation path: extend @claude-flow/memory with a pool manager; measure vs current per-agent baseline. Priority: high — this is a Grade A finding with 97.7% measured reduction.

  2. Implement disaggregated prefill/decode scheduling (ConServe pattern) — Add a serving-layer scheduler that separates prefill (prompt processing) from decode (token generation) across the agent pool. Expected: 50%+ p95 TTFT reduction. Implementation path: new PerformanceScheduler module in @claude-flow/cli; gate behind CLAUDE_FLOW_SCHED_DISAGGREGATE=1 flag.

  3. Agent-to-agent channel security audit against MESA threat model — Run AI-Infra-Guard's 75+ rules against Ruflo's MCP server pipelines and SendMessage pathways. A single compromised edge = 75% attack success. Immediate action: label all agent-to-agent edges as attack surface in @claude-flow/security threat model; add static taint gate to pre-task hook.