Unverified50% confidenceFactExact time
Long-horizon consistency refers to the model's ability to keep goals, state, and logic from drifting after spanning dozens of decision steps, which is not the same thing as context window size.
1
Sources
50%
Confidence
Long-term
Relevance
11/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedSmall models at the 9B level are prone to goal drift in long-chain, multi-step autonomous task execution due to limitations in effective context window utilization and multi-step reasoning stability72% similarUnverified上下文回贯机制缺失导致模型每轮接收到的上下文完全一致，从而机械重复相同决策形成无限循环69% similarVerifiedIdempotency means that making multiple calls to the same operation produces exactly the same result as making it once; a unique idempotency key should be generated for each business action and the result persistently bound to that key to prevent duplicate execution.69% similarUnverifiedIndividually, deformation fields, anchor coordination, zero-displacement prediction, size-based grouping, and symmetry enhancement are not novel, but together they form a deployable pipeline.68% similarUnverifiedThe goal mode is most valuable for repetitive work with clear completion states, while the instruction mode is better suited for one-off tasks and novel contexts.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1123105API
curl https://kongchang.com/api/v1/knowledge/claims/1123105MCP
get_claim(id=1123105)