Johns Hopkins Open-Sources Gricea, a Platform Replicating 93% of Conversational AI Study Configs
JohnsHopkins · hf · 2026-09-21
Johns Hopkins researchers present Gricea, an open-science platform for conversational AI (CAI) research that tackles fragmented, hard-to-replicate study setups.
- Studies are represented as configurable, deployable research artifacts coupling procedures, participant-facing systems, and conversational task behavior — runnable, inspectable, shareable, and reusable
- A replication study reproduced 93% of eligible CUI 2026 paper configs while flagging missing information in 96% of papers that hinders faithful replication
- A user study showed researchers and practitioners from diverse backgrounds could build runnable studies addressing open-ended questions
Gricea enables cumulative knowledge building in CAI research through shared artifacts.
Related event: Johns Hopkins Open-Sources Gricea to Tackle AI Research Reproducibility(2 posts)→
More from Research
- Frontier Data Summit 2026 lineup reveals a dozen new AI benchmarks and top researchers — dlwh · 2026-09-22
- RL run spends a lot of compute on graders; is re-prefilling worth it over 30 steps — stochasticchasm · 2026-09-22
- Diag2Diag: AI generates measurements that hardware sensors can't capture — AnneliesGamble · 2026-09-22
- Overnight JEV-style model run scores just 24% on 120 hard tasks — BLUECOW009 · 2026-09-22
- New psychology paper uses social identity to explain false beliefs in AI-era information environments — steverathje2 · 2026-09-22
- Standard GRPO at 1M scale: why no critic models, and what counts as "behaviors"? — stochasticchasm · 2026-09-22