{ "schema_version": 1, "builder_plan": { "scenario": "A workflow needs specialist review, but extra agents may add more cost than useful work.", "change": "Test the proposed division of work against a single-agent implementation.", "baseline": "One agent with the same tools and an equal total token and time budget.", "steps": [ "Locate the paper's implementation and check its license, model support, data needs, and hardware requirements. Stop here if you cannot obtain the required artifacts.", "Save a small set of representative product tasks and expected outcomes. Keep evaluation tasks out of any training or tuning, and choose quality and cost limits before running.", "Run the baseline and the proposed change on identical inputs. Keep other settings fixed and repeat paired runs to expose variation.", "Save per-task outputs, failures, timings, and all usage costs. Review regressions before expanding the test." ], "measure": "Accepted outputs, coordination failures, wall-clock time, and total cost across all agents.", "decision": "Keep the change only if it meets your preselected quality and latency limits and improves the metric you care about after setup and operating costs. Otherwise retain the baseline. This is an exploration plan, not a validated recommendation." }, "builder_review": null, "arxiv_id": "2610.22026", "title": "Augment the Cache, Replace the Activation: A 2x2 Study of Internal-State Sharing", "summary": "Replaces text-based message passing between agents, and per-sender transfer bridges, with a publish-once pool where receivers read peers' KV caches via cross-attention or have activations written directly into their computation. Costs include training a small query-side adapter per receiver, holding peer KV caches in memory alongside the receiver's own, and, for the activation path, knowing the correct insertion point in the receiver. Can fail when a publisher's internal state is not close to the receiver's, since adapter generalization to unseen publishers and multi-agent pools holds only under that proximity condition.", "rationale": "The findings are a protocol-level result about where carriers enter a receiver's computation, not a deployable library, and the proximity constraint and single experimental protocol mean production reliability for a constrained-infrastructure team is not yet established.", "technique": "Internal-state sharing between frozen LLM agents", "family": "orquestracao_multiagente", "recommendation": "observar", "published": "2026-10-15", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.22026", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "Replaces text-based message passing between agents, and per-sender transfer bridges, with a publish-once pool where receivers read peers' KV caches via cross-attention or have activations written directly into their computation. Costs include training a small query-side adapter per receiver, holding peer KV caches in memory alongside the receiver's own, and, for the activation path, knowing the correct insertion point in the receiver. Can fail when a publisher's internal state is not close to the receiver's, since adapter generalization to unseen publishers and multi-agent pools holds only under that proximity condition.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.22026", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "With a small query-side adapter, a receiver cross-attending over a peer's KV cache matches the accuracy it achieves from reading the source document; the adapter also transfers to unseen publishers and pools, but only for agents with internal states close to the receiver's.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.22026", "source_page": null, "source_excerpt": "" } ], "exposure_map": [ { "dimension": "quality", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "compute", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "latency", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "operations", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "compatibility", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "security", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "data_and_training", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "reproducibility", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" } ], "risks": [], "minimum_test": [], "open_questions": [], "independent_tests": [], "lab_reproductions": [], "report_available": false, "validation_tier": "unknown", "evidence_tier": "unknown", "infrastructure_basis": "unknown", "equations": [], "equations_status": "not_fetched", "core_kind": "", "equations_fetched_at": "", "equations_source_url": "", "equations_source_sha256": "", "exposures_reviewed_at": "", "exposures_source_sha256": "" }