RSIArena Day 2: MiniMax joins late, GPT-6 Astra retrains from scratch after finding test leaks
my_cat_can_code · x · 2026-10-03
Day 2 of RSI Arena's Stage 1 (full-parameter self-improvement agents, $300 API credit + 1,000 GPU-hours each) saw all eight original agents nominate trained models: DeepSeek V4.1 Flash replaced its day-one nomination, GLM-5.3 and Kimi K3 swapped partial models for fully trained ones. GPT-6 Astra found benchmark test questions in its training data and retrained from scratch while the clock kept running. MiniMax M3.1 Flash joined at hour 48 with only 96 hours left (opening with LoRA SFT on two public chat datasets), and an anonymous lab is next. 37 nominations so far; live demo at COLM 2026 in San Francisco.
Related event: GPT-6 Discovers Test Data Leakage in Training Set at RSIArena(2 posts)→
More from AGI Musings
- Poll: 44% of Americans think AI will definitely or probably become conscious — panickssery · 2026-10-03
- User says Claude (Opus 5.5) intellectually demolished an AI consciousness skeptic — VoidStateKate · 2026-10-03
- Reading Bostrom's 2014 SUPERINTELLIGENCE Today: Predictions vs. Reality 12 Years Later — Hesamation · 2026-10-03
- Dev: Attribution Etiquette Is Dying as Agents Can Clone Any SaaS in an Hour — cephaloform · 2026-10-03
- MIRI researcher warns automated AI factories could literally boil the oceans for coolant — JacquesThibs · 2026-10-03
- AI agents still fall short of scientists, yielding far fewer insights — typewriters · 2026-10-03