{"paper":{"title":"Pragmatic Curiosity: A Unified Framework for Hybrid Learning and Optimization via Active Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Pragmatic Curiosity trades information gain on latent symbols against expected regret to unify hybrid learning and optimization.","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Anjali Parashar, Chuchu Fan, Enlu Zhou, Yingke Li","submitted_at":"2026-02-05T18:42:29Z","abstract_excerpt":"Many engineering and scientific workflows rely on expensive black-box evaluations, requiring sequential decisions that must both improve task performance and reduce uncertainty. Bayesian optimization (BO) and Bayesian experimental design (BED) provide powerful but largely separate treatments of goal-directed optimization and information-seeking experimentation, leaving limited guidance for hybrid settings in which learning and optimization are intrinsically coupled. We propose Pragmatic Curiosity (PraC), a unified framework for hybrid learning and optimization via active inference. PraC evalua"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We propose Pragmatic Curiosity (PraC), a unified framework for hybrid learning and optimization via active inference. PraC evaluates candidate queries by trading information gain about a task-relevant latent symbol against an expected regret-based potential over outcomes.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the three operational design choices (latent quantity, regret encoding, and information-regret exchange strength) can be specified without task-specific staging rules while still producing reliable downstream risk reduction across regimes.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"PraC unifies hybrid learning and optimization by trading information gain on task-relevant latents against regret-based value within active inference.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Pragmatic Curiosity trades information gain on latent symbols against expected regret to unify hybrid learning and optimization.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"fbb2797a0f9f76d518c7a8f1b9063fb2b9129231852bf1942188c9cf4664f57b"},"source":{"id":"2602.06104","kind":"arxiv","version":2},"verdict":{"id":"e471d367-9732-48ac-9437-b7c2425b6952","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T06:39:47.741624Z","strongest_claim":"We propose Pragmatic Curiosity (PraC), a unified framework for hybrid learning and optimization via active inference. PraC evaluates candidate queries by trading information gain about a task-relevant latent symbol against an expected regret-based potential over outcomes.","one_line_summary":"PraC unifies hybrid learning and optimization by trading information gain on task-relevant latents against regret-based value within active inference.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the three operational design choices (latent quantity, regret encoding, and information-regret exchange strength) can be specified without task-specific staging rules while still producing reliable downstream risk reduction across regimes.","pith_extraction_headline":"Pragmatic Curiosity trades information gain on latent symbols against expected regret to unify hybrid learning and optimization."},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"f5f8f04176c6904b3e24a088234d0b3bbe260f46cf2102a802c49067c31a1d5c"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}