LEVI: Stronger Search Architectures Can Substitute for Larger LLMs in Evolutionary Search

LEVI: Stronger Search Architectures Can Substitute for Larger LLMs in Evolutionary Search | Signal Canvas | ScienceToStartup

Page Freshness

Signal Canvas proof surface

Canonical route: /signal-canvas/levi-stronger-search-architectures-can-substitute-for-larger-llms-in-evolutionary-search

stale

Proof freshness: fresh
Proof status: unverified
Display score: 9/10
Last proof check: 2026-05-12
Score updated: 2026-05-12
Score fresh until: 2026-06-11
References: 0
Source count: 0
Coverage: 0%

This page is showing the last landed evidence receipt and score bundle because the latest proof data is outside the freshness window.

Agent Handoff

Canonical ID levi-stronger-search-architectures-can-substitute-for-larger-llms-in-evolutionary-search | Route /signal-canvas/levi-stronger-search-architectures-can-substitute-for-larger-llms-in-evolutionary-search

REST example

curl https://sciencetostartup.com/api/v1/agent-handoff/signal-canvas/levi-stronger-search-architectures-can-substitute-for-larger-llms-in-evolutionary-search

MCP example

{
  "tool": "search_signal_canvas",
  "arguments": {
    "mode": "paper",
    "paper_ref": "levi-stronger-search-architectures-can-substitute-for-larger-llms-in-evolutionary-search",
    "query_text": "Summarize LEVI: Stronger Search Architectures Can Substitute for Larger LLMs in Evolutionary Search"
  }
}

source_context

{
  "surface": "signal_canvas",
  "mode": "paper",
  "query": "LEVI: Stronger Search Architectures Can Substitute for Larger LLMs in Evolutionary Search",
  "normalized_query": "2605.09764",
  "route": "/signal-canvas/levi-stronger-search-architectures-can-substitute-for-larger-llms-in-evolutionary-search",
  "paper_ref": "levi-stronger-search-architectures-can-substitute-for-larger-llms-in-evolutionary-search",
  "topic_slug": null,
  "benchmark_ref": null,
  "dataset_ref": null
}

Evidence Receipt

Route status: building

Claims: 12

References: Pending verification

Proof: Verification pending

Freshness state: computing

Source paper: LEVI: Stronger Search Architectures Can Substitute for Larger LLMs in Evolutionary Search

PDF: https://arxiv.org/pdf/2605.09764v1

Repository: https://github.com/ttanv/levi

Source count: Pending verification

Coverage: 0%

Last proof check: 2026-05-12T20:14:59.855Z

Signal Canvas receipt window

Ready for execution: LEVI: Stronger Search Architectures Can Substitute for Larger LLMs in Evolutionary Search

/buildability/levi-stronger-search-architectures-can-substitute-for-larger-llms-in-evolutionary-search

Build Nowready

Subject: LEVI: Stronger Search Architectures Can Substitute for Larger LLMs in Evolutionary Search

Verdict

Build Now

Preparing verified analysis

GitHub Code Pulse

Cached

Stars

Health

Last commit

6/7/2026

Forks

Open repository

Claim map

Strong 12Mixed 0Weak 0

Evidencepartial
Across systems-research benchmarks LEVI attains the highest score on a budget 3.3-6.7x smaller than the published frontier-model runs of existing frameworks like ShinkaEvolve, GEPA, and AdaEvolve
Implicationpartial
Directly stated in abstract with specific numeric comparison.
Verificationpartial
partial
Evidencepartial
on one problem, LEVI matches the existing best at a 35x lower cost
Implicationpartial
Directly stated in abstract with specific numeric claim.
Verificationpartial
partial
Evidencepartial
On prompt optimization, LEVI matches or exceeds GEPA at less than half of its rollout budget on four different benchmarks
Implicationpartial
Directly stated in abstract with specific comparison.
Verificationpartial
partial
Evidencepartial
LEVI improves on three core components of evolutionary search: a solution database that establishes diversity from the beginning, and then maintains it throughout the run; a smarter mutation router that plays into the strengths of large and small LLMs; and a rank-preserving proxy benchmark for rollout-heavy settings
Implicationpartial
Directly stated in abstract as the key methodological contributions.
Verificationpartial
partial
Evidencepartial
archives that fail to preserve solution diversity force compensation through stronger mutation models
Implicationpartial
Stated as a problem that LEVI addresses, implying it is a limitation of prior work.
Verificationpartial
partial
Evidencepartial
blind model use spends frontier dollars on local edits a smaller model could handle
Implicationpartial
Stated as a problem that LEVI addresses, implying inefficiency in prior work.
Verificationpartial
partial
Evidencepartial
full-set evaluation wastes rollouts on redundant examples
Implicationpartial
Stated as a problem that LEVI addresses, implying inefficiency in prior work.
Verificationpartial
partial
Evidencepartial
LEVI is available as an open-source framework at https://github.com/ttanv/levi
Implicationpartial
Directly stated in abstract with a URL.
Verificationpartial
partial
Evidencepartial
Across systems-research benchmarks LEVI attains the highest score on a budget 3.3-6.7x smaller than the published frontier-model runs of existing frameworks like ShinkaEvolve, GEPA, and AdaEvolve
Implicationpartial
Directly stated in abstract with specific numeric range and named baselines.
Verificationpartial
partial
Evidencepartial
on one problem, LEVI matches the existing best at a 35x lower cost.
Implicationpartial
Directly stated in abstract with specific multiplier.
Verificationpartial
partial
Evidencepartial
On prompt optimization, LEVI matches or exceeds GEPA at less than half of its rollout budget on four different benchmarks.
Implicationpartial
Directly stated in abstract with specific comparison and number of benchmarks.
Verificationpartial
partial
Evidencepartial
LEVI improves on three core components of evolutionary search: a solution database that establishes diversity from the beginning, and then maintains it throughout the run; a smarter mutation router that plays into the strengths of large and small LLMs; and a rank-preserving proxy benchmark for rollout-heavy settings.
Implicationpartial
Directly stated in abstract as the key methodological contributions.
Verificationpartial
partial

Author intelligence and commercialization panels stay hidden until the proof receipt is verified, cites at least 3 references, includes at least 2 sources, and clears 50% coverage. The paper narrative and citation surfaces remain public while verification is pending.

LEVI: Stronger Search Architectures Can Substitute for Larger LLMs in Evolutionary Search

Use Signal Canvas as the narrative proof surface

Use this Signal Canvas via API or MCP

Signal Canvas proof surface