Unverified50% confidenceEventExact time
The first version of the evaluation used GPT-4 as a judge to read the conversation trajectory and output noticed: true/false, but it collapsed in actual use for three specific reasons: position bias in long conversations, the judge model's tendency toward charitable interpretation, and non-determinism across runs.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/15/2026
First Seen
Valid until: 1/13/2027
Sources
Related Entities
Related Claims
UnverifiedGPT-4等模型在作为评判者时会系统性地偏好特定位置的回答77% similarUnverified评分模型（如GPT-4）在版本更新后，其对优质回答的判断标准可能发生偏移，导致历史评分数据出现系统性断层74% similarUnverifiedThe technical feasibility of voice-to-intent products is largely underpinned by the emergent capabilities of models such as GPT-4, as rules alone or small models previously could not stably perform cross-context text rewriting.74% similarUnverifiedGPT-4 and subsequent models are already able to provide more honest feedback and constructive alternative suggestions when faced with clearly unreasonable requests.73% similarUnverifiedEdu-QuRating uses GPT-4.1-mini as the judge to make pairwise preference judgments on document pairs.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1036220API
curl https://kongchang.com/api/v1/knowledge/claims/1036220MCP
get_claim(id=1036220)