# How Mira Matches Candidates Source: https://docs.mira.day/en/docs/match-quality-and-trust/understanding-ai-matching > Why Mira reasons about your requirements instead of matching keywords, how it scores each candidate against your criteria with evidence, and where the technology comes from. On most platforms, searching for candidates means keyword matching: you type "Python developer Berlin" and get everyone with those words on their profile. It's fast, but it misses a lot, because the best candidate rarely describes themselves in your exact words. Mira works differently. Mira Reasoning Embedding (MRE), its reasoning-based retrieval model, reads your requirements, reasons about what you actually need, and surfaces the candidates whose real career history fits. Mira then scores each one against your criteria. The examples below show what that looks like in practice. *** What reasoning catches that keywords miss [#what-reasoning-catches-that-keywords-miss] * **"5 years of backend experience."** Keyword search looks for "5 years" in the text. Mira adds up the actual time from the candidate's job history, so it counts 3 years at one company plus 2.5 at another as backend experience, even if the profile never says "5 years." * **"Someone who's built a product from 0 to 1."** Keyword search looks for the literal phrase. Mira recognizes the pattern: joining a company early and staying through its growth. * **"High-growth company background."** Keyword search can't do this at all. Mira reads through the candidate's work history and reasons about the company's stage and trajectory. * **"Public company plus startup experience."** Keyword search treats these as two unrelated filters. Mira reads through the whole career arc and finds people who did both at different points, which signals a specific kind of versatility. Describe roles the way you'd tell a colleague [#describe-roles-the-way-youd-tell-a-colleague] Because Mira reasons, you can describe a role in plain language: > *"I need someone senior, technical but customer-facing, who's done enterprise sales at a startup that actually had product-market fit, not one that just raised a lot of money."* Mira unpacks that into separate criteria and applies each one. No keyword combinations, filters, or search syntax. How Mira scores each candidate [#how-mira-scores-each-candidate] Mira doesn't hand back a black-box score. For each candidate, it goes through your criteria one at a time and marks each as **met**, **uncertain**, or **not met**, with a line of evidence pulled from that candidate's own history. When a profile simply doesn't say, Mira marks it uncertain rather than guessing. From those per-criterion results, each candidate gets an overall label, **Good Match**, **Partial Match**, or **Low Match**, so you can scan the pool fast and still see the reasoning behind every call. There are no numeric scores to decode. Where the technology comes from [#where-the-technology-comes-from] MRE comes out of Mira's own research into reasoning-based retrieval. As of July 2026, it ranks first overall on BRIGHT, a public benchmark for reasoning retrieval, and first among embedding-based retrievers. Mira also built PJBenchmark, a recruiting-specific evaluation set built on real hiring data. Four papers describe the approach: * [Time-Scaling Is What Agents Need Now](https://arxiv.org/abs/2601.02714) * [Refine Thought](https://arxiv.org/abs/2511.13726) * [PJB](https://arxiv.org/abs/2603.17386) * [CRE-T1 Preview Technical Report](https://arxiv.org/abs/2603.17387) What it doesn't do [#what-it-doesnt-do] Mira won't invent a candidate who isn't there, and it doesn't make the hiring decision, it surfaces the best available matches with the evidence, and the final call is always yours. For the full picture of what to rely on Mira for and what stays with you, see [What Mira Can & Can't Do](/en/docs/match-quality-and-trust/what-mira-can-and-cant-do). Getting better results [#getting-better-results] The quality of matching depends on the quality of your input: * Be specific, and you get sharper matches. * Include implicit criteria, like company stage or career pattern, to give Mira more to reason about. * Refine after seeing results, each round gets sharper. See [How Mira Searches](/en/docs/sourcing/agentic-search) for practical tips.