Unverified50% confidenceFactExact time
Testing revealed that asking the model to reason two more steps ahead from the same board position does not necessarily yield better results, and sometimes the final move is actually worse.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/9/2026
First Seen
Valid until: 1/7/2027
Sources
Related Entities
Related Claims
UnverifiedIn experiments where the chain was progressively lengthened, all three types of models were correct at one step, but from the second step onward the two 'one-glance models' began making errors, and after a few more steps were almost entirely wrong, while the thinking model performed near-perfectly throughout with only its token count increasing.78% similarUnverifiedBenchmark saturation refers to the point at which, after model performance improves to a certain level, benchmark tests can no longer effectively differentiate capability differences between models.74% similarUnverifiedSelf-consistency checking, by having the model sample the same question multiple times and comparing answer stability, consumes only inference time without increasing peak VRAM usage.72% similarUnverifiedIf two safety benchmarks use different harnesses, simply comparing their scores becomes meaningless, because what is being compared are two complete systems rather than two models.71% similarUnverifiedAs model capabilities improve, the tension between capability and alignment becomes increasingly pronounced, and the consequences of alignment failure are correspondingly amplified.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/993097API
curl https://kongchang.com/api/v1/knowledge/claims/993097MCP
get_claim(id=993097)