[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-gacha-decoding-makes-smarter-ai-models-more-creative-not-less":10,"sections":34},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":29,"feedback":33,"feedback_at":22,"cost_usd":33,"total_tokens":33},9529,"gacha-decoding-makes-smarter-ai-models-more-creative-not-less","Gacha Decoding Makes Smarter AI Models More Creative, Not Less","A new AI sampling method uses randomness and instruction-following to generate more diverse outputs, reversing the usual diversity-capability tradeoff.","Researchers have a new way to make AI models give more varied answers: have the model follow dice-roll instructions instead of just guessing randomly.\n\nA new research paper posted October 2, 2026 introduces a technique called Gacha Decoding. Instead of relying on a language model's built-in randomness (how unpredictable its next word choice is) to generate different answers, Gacha pairs the model's ability to follow instructions with an external random-number generator - a kind of \"planning with dice.\" Tested across chatbot conversations, creative writing, image-generation planning, and protein design, the method produced far more distinct high-quality answers than existing diversity techniques: up to 2.4 times higher diversity score (Vendi score) at equal quality, and it needed 11 times fewer samples to reach the same number of good, distinct answers. It also surfaced entirely new types of answers that other methods missed.\n\nThat matters because of a trend researchers keep running into: as language models get better at following instructions, they tend to get more confident and predictable - great for accuracy, bad for variety. Gacha Decoding flips that pattern. Diversity under this method goes up as the model gets more capable, even while the model's own token-level randomness keeps shrinking.\n\nThe name fits: it treats answer generation less like a coin flip and more like a loot-box pull, with randomness injected deliberately instead of left to the model's internal fuzziness. It's one not-yet-peer-reviewed paper, so treat those multipliers as the authors' own numbers until someone else reproduces them.","[\"ai\",\"llm-decoding\",\"generative-ai\",\"research\"]","2026-10-02T04:00:00.000Z","2026-10-02T23:22:15.374Z","2026-10-02T23:22:21.885Z","published",null,[],"ai",[24,26,27,28],"llm-decoding","generative-ai","research",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.01382",0,{"sections":35},[36,39,43,47,52,56,60,65,70,75,80,85,90,95],{"name":37,"slug":24,"count":38,"latest_published_at":18},"AI",5896,{"name":40,"slug":41,"count":42,"latest_published_at":18},"Security","security",837,{"name":44,"slug":45,"count":46,"latest_published_at":18},"Policy","policy",438,{"name":48,"slug":49,"count":50,"latest_published_at":51},"Deals","deals",317,"2026-10-01T22:00:00.000Z",{"name":53,"slug":54,"count":55,"latest_published_at":18},"Hardware","hardware",199,{"name":57,"slug":58,"count":59,"latest_published_at":18},"Science","science",171,{"name":61,"slug":62,"count":63,"latest_published_at":64},"Consumer Tech","consumer-tech",155,"2026-10-01T19:54:10.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Dev Tools","dev-tools",96,"2026-10-01T16:57:03.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":74},"Software","software",93,"2026-09-30T21:41:11.000Z",{"name":76,"slug":77,"count":78,"latest_published_at":79},"Startups","startups",90,"2026-10-01T21:55:22.000Z",{"name":81,"slug":82,"count":83,"latest_published_at":84},"Gaming","gaming",53,"2026-10-02T02:50:39.000Z",{"name":86,"slug":87,"count":88,"latest_published_at":89},"General","general",50,"2026-09-30T21:37:54.000Z",{"name":91,"slug":92,"count":93,"latest_published_at":94},"Reviews","reviews",31,"2026-09-28T14:31:34.000Z",{"name":96,"slug":97,"count":98,"latest_published_at":99},"How-To","how-to",7,"2026-10-01T09:00:00.000Z"]