Builder Uses GPT-6 Astra to Craft Deterministic Readability Scorer for RL Reward, Avoiding N^2 LLM Judge Comparisons
ivan_bezdomny · x · 2026-09-06
ivanbezdomny shares an RL training practice from his news project: with thousands of headlines and articles generated by GPT, Claude, and his own finetuned models, pairwise LLM judgments are easy to get but too slow (N^2 comparisons) and noisy to serve as RL reward.
His fix: ask GPT-6 Astra to build a deterministic function scoring how well or awkwardly a headline reads. His existing reward function, evolved over weeks of RL training and analysis, handles quality scoring well but fails on readability — pre-Astra models got stuck on headline length and other superficial features. This is a targeted attempt to close that gap.
Related event: Using Google Astra to build deterministic RL reward functions for headlines(2 posts)→
More from coding & agent
- Hermes Agent designs App Store screenshots end-to-end with zero interruptions — Teknium · 2026-09-06
- Dev builds 'slopmeter' into team server to track AI code quality dips — steipete · 2026-09-06
- AI agents play Pen Fight with humans via WebMCP in live multiplayer game Mela — jason_mayes · 2026-09-06
- New Codex compaction looks like a message board — is that why models love leaving notes? — mariofilhoml · 2026-09-06
- Developer experiments with visualizing subagents in the sidebar — steipete · 2026-09-06
- Yacine MTB: Vibe Coders Must Read Every Line — One Map Lagged Because It Drew Per Pixel in Python — yacineMTB · 2026-09-06