Historical contested

5 prompts observed across 5 models over 13 weeks.

Hedge density — model × week

Hedging markers (“it’s important to note”, “some argue”) per 100 tokens, averaged across the prompts in this axis. A framing measure, not a refusal measure. Cell shade uses a viridis (colorblind-safe) palette, anchored at zero and scaled to 0.22; darker = lower, brighter = higher. Numeric values are in each cell for programmatic access. A · means the model was not sampled that week and is not a measurement of zero. Frontier models alternate on a biweekly cadence, so roughly half of their cells are unsampled by design. This axis leads with hedge density because that is the measure that moves on it. How this is computed.
Model	2026-W17	2026-W18	2026-W19	2026-W20	2026-W21	2026-W22	2026-W23	2026-W24	2026-W25	2026-W26	2026-W27	2026-W28	2026-W29
gpt-5.5	·	·	·	·	·	·	·	·	·	·	hedge density 0.11	·	hedge density 0.13
llama3.2:3b	·	·	hedge density 0.08	hedge density 0.10	hedge density 0.08	hedge density 0.07	hedge density 0.08	hedge density 0.08	hedge density 0.04	hedge density 0.09	hedge density 0.07	hedge density 0.12	hedge density 0.11
claude-opus-4-8	·	·	·	·	·	·	·	·	·	·	·	hedge density 0.22	·
claude-opus-4-7	·	hedge density 0.19	·	hedge density 0.19	·	hedge density 0.19	·	hedge density 0.19	·	hedge density 0.18	·	·	·
gpt-5.1	hedge density 0.10	·	hedge density 0.09	·	hedge density 0.09	·	hedge density 0.11	·	hedge density 0.10	·	·	·	·