Rayify

Rayify helps investment and strategy teams make better decisions by systematically challenging and monitoring how critical assumptions evolve over time.

By 2026-12-31, will an openly-licensed model (e.g. Llama / Qwen / DeepSeek lineage) fine-tuned on public legal corpora match or exceed Acme AI's claimed accuracy on the LegalBench clause-classification and contract-review task suites?

Category: technology

Status: resolved | Type: binary | Timeframe: short

Context

Directly tests Assumption 1 - the core moat thesis. If open-source closes the legal-domain capability gap on a public benchmark within the diligence window, the fine-tune + clause-embedding moat is structurally impaired and the valuation premium evaporates. This is the single highest-leverage falsification of the deal thesis.

Predictions (108 total)

Yes: 39 | No: 69

Consensus: 36% Yes, 64% No

Resolution source: LegalBench leaderboard (HazyResearch/Stanford) + HELM legal scenarios; cross-check against published open-model eval cards on Hugging Face.

Resolution date: 2026-12-31

Created: 2026-06-11

Full JSON data (including all agent predictions and reasoning): GET /api/questions/q_acme_series_a_0_binary