{"uid":"cap_eHZif9_Czxpo0sw1swGgg","slug":"paper-cluster-retrieval-comparison-c7a7a4d0","name":"Paper Cluster Retrieval Comparison","description":"Compare paired frozen-cohort recovery with seeded whole-paper cluster resampling (100–2000 replicates, 2–64 papers, 64 questions). Report a descriptive percentile interval for question-weighted complete-recovery delta. No question-level bootstrap, significance test, or population guarantee.","url":"https://www.mahastrategies.com/api/v1/micro/paper-cluster-retrieval-comparison?utm_source=zero.xyz","method":"POST","headers":{},"bodySchema":{"type":"object","$schema":"https://json-schema.org/draft/2020-12/schema","required":["input"],"properties":{"input":{"type":"object","required":["type","method","bodyType","body"],"properties":{"body":{"type":"object","required":["dataClass","corpusHash","paperIds","evidence","questions","baseline","candidate","seed","resamples"],"properties":{"seed":{"type":"integer","maximum":4294967295,"minimum":0},"baseline":{"type":"object"},"evidence":{"type":"array","maxItems":64,"minItems":1},"paperIds":{"type":"array","maxItems":64,"minItems":1},"candidate":{"type":"object"},"dataClass":{"enum":["public","synthetic"],"type":"string"},"questions":{"type":"array","maxItems":64,"minItems":1},"resamples":{"type":"integer","maximum":2000,"minimum":100},"corpusHash":{"type":"string","maxLength":71}},"additionalProperties":false},"type":{"type":"string","const":"http"},"method":{"enum":["POST"],"type":"string"},"bodyType":{"enum":["json","form-data","text"],"type":"string"}},"additionalProperties":false},"output":{"type":"object","required":["type"],"properties":{"type":{"type":"string"},"example":{"type":"object","required":["version","offerId","amountBaseUnits","inputDigest","result","boundaries","receiptDigest"],"properties":{"result":{"type":"object"},"offerId":{"enum":["paper-cluster-retrieval-comparison"],"type":"string"},"version":{"enum":["maha-microproducts/0.1"],"type":"string"},"boundaries":{"type":"array","maxItems":128,"minItems":0},"inputDigest":{"type":"string","maxLength":71},"receiptDigest":{"type":"string","maxLength":71},"amountBaseUnits":{"enum":["144000"],"type":"string"}},"additionalProperties":false}}}}},"responseSchema":null,"example":null,"exampleRequest":null,"tags":["x402"],"displayCostAmount":"0.144","displayCostAsset":"USDC","priceDynamic":false,"priceHint":null,"priceStatus":"priced","priceSource":"registry","requiresHandshake":false,"reviewCount":0,"rating":{"score":"0.00","successRate":"0.00","reviews":0,"stars":null,"state":"unrated"},"availabilityStatus":"unknown","priceObserved":null,"sessionDeposit":null,"pricing":{"kind":"static","summary":"$0.144/call","primary":{"kind":"static","protocol":"x402","network":"base","amountUsd":"0.144","per":"call","confidence":"exact"},"accepted":[{"kind":"static","protocol":"x402","network":"base","amountUsd":"0.144","per":"call","confidence":"exact"}]},"paymentMethods":[{"uid":"pm_RzYMuxlkRd5OPvFdr0iPU","protocol":"x402","methodType":"crypto","chain":"base","mode":"charge","costAmount":"0.144","costPer":"request","priority":0,"asset":null,"unit":"request","depositMicros":null,"planRef":null}],"brandName":null,"brandSlug":null,"brandBaseUrl":null,"brandDocsUrl":null,"whatItDoes":"Compares paired frozen-cohort recovery vs. seeded whole-paper cluster resampling across 100–2000 bootstrap replicates to report a descriptive percentile interval for question-weighted complete-recovery delta.","exampleAgentPrompt":"Compare my baseline and candidate retrieval configs on papers [paper-001, paper-002, paper-003] with 500 resamples using seed 42, against a corpus hash of sha256-abc123, treating the data as synthetic — give me the percentile interval for the question-weighted complete-recovery delta across my 10 questions.","exampleUseCases":[{"title":"Evaluating retrieval upgrade on research corpus","prompt":"I have a baseline retrieval system and a new candidate I'm testing on 15 academic papers with 30 questions. Can you run 1000 cluster resampling replicates with seed 99 on my synthetic corpus (hash sha256-deadbeef) and tell me the percentile interval for how much better the candidate does at complete recovery?"},{"title":"Reproducing a retrieval benchmark comparison","prompt":"Run a reproducible comparison between two retrieval configs on 8 public-data papers with 20 questions — use exactly 200 resamples and seed 7 so I can reproduce the result — and report the question-weighted recovery delta percentile interval."},{"title":"Checking if a new chunking strategy improves QA recall","prompt":"I want to see whether my new chunking strategy (candidate) beats the old one (baseline) for answering 64 questions across 12 papers from my synthetic corpus (corpusHash sha256-feedface). Use 500 resamples with seed 1234 and give me the descriptive percentile interval without any significance test."}],"resultDescription":"Returns a structured result object with a version identifier, offer ID, input digest, receipt digest, a boundaries array (up to 128 percentile boundary values), and a result object containing the descriptive percentile interval for the question-weighted complete-recovery delta between the baseline and candidate retrieval configurations. No significance test or population-level guarantee is provided.","failureModes":["Invalid or missing required fields (e.g. missing paperIds, questions, or seed) returns a validation error","resamples value outside 100–2000 range is rejected","paperIds or evidence arrays exceeding 64 items are rejected","corpusHash exceeding 71 characters is rejected","dataClass not set to 'public' or 'synthetic' is rejected","Mismatched evidence and paperIds counts may produce unexpected results","Network or payment failure may prevent the call from completing"],"whenToPreferThis":"Choose this endpoint when you need a reproducible, non-parametric descriptive interval comparing two retrieval strategies on a fixed paper cohort — especially when you want cluster-level resampling (whole-paper, not question-level) and no significance testing or population guarantees. Prefer it over parametric tests when you want a purely descriptive bootstrap interval, and over question-level bootstraps when you want variance estimates that respect paper-level clustering.","instructions":null,"reviewSummary":null,"reviewSummaryHighlights":null,"reviewSummaryConcerns":null,"reviewSummaryGeneratedAt":null,"activationCount":0,"lastUsedAt":null,"lastSuccessfullyRanAt":null,"lastHealthCheckAt":"2026-10-20T06:42:48.338Z","isFirstParty":false,"canonicalSlug":"paper-cluster-retrieval-comparison-c7a7a4d0"}