Google open-sources Agile Deliberation, an AI framework that turns fuzzy classification rules into explicit ones
aigclink · x · 2026-11-06
Google Research open-sourced Agile Deliberation, a human-in-the-loop framework for subjective visual classification where there's no single right answer (e.g., what counts as "healthy food").
The key inversion: instead of humans telling AI the rules, AI actively helps humans discover their own criteria.
- Concept Scoping: decomposes the vague concept via prompt chaining, generating positive/negative categories into a structured In-Scope/Out-of-Scope definition
- Concept Iteration: the system mines borderline images (dedup/clustering via dictionary learning, cluster selection via bandit algorithms, ambiguity mining), has users label them, then greedily selects the best rule refinement from feedback
The pipeline ends with ready-to-use image classifiers. Use cases include content moderation (defining misleading ads) and enterprise data labeling (consistent quality/brand-style standards). Code and a step-by-step Demo notebook are on the google-research GitHub repo.
Related event: Google Open-Sources Agile Deliberation for Ambiguous Classification(4 posts)→
More from coding & agent
- Chalk talk: Jediah Katz gives an inside look at their agent harness design — jxnlco · 2026-11-06
- Tinker team benchmarks model finetunability: Inkling 2 Small learns faster for less — ziqiao_ma · 2026-11-06
- 28,849-trial Harness Leaderboard: bash-only mini-swe-agent beats openclaw 51.4% vs 27.3% — bansalg_ · 2026-11-06
- Qwen 27B + Python REPL harness hits 78.9% on Harvey benchmark, beating larger flagships — bansalg_ · 2026-11-06
- One Slack ask, 34 merged PRs in under 6 hours: Sentry CEO on running coding agents — zeeg · 2026-11-06
- Agents Beat Hidden-Info Eval by Brute-Forcing RNG Seed 2026 to Recover Pixel Permutation — RobertTLange · 2026-11-06