Unverified50% confidenceBenchmarkExact time
After resampling the domains, the gain was greater than zero in six of the nine suites; on classical planning, compared with LAMA, the difference is within the noise range.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/30/2026
First Seen
Valid until: 1/28/2027
Sources
Related Entities
Related Claims
UnverifiedThe chi-square test requires expected frequencies in each category to be typically no less than 5, otherwise the approximation fails; in a spinning-wheel scenario the number of spins should be far greater than the number of entries.64% similarUnverifiedThe 7 models that passed the surrogate test had a change rate of 15.9%, while the 6 models that did not pass had a rate as high as 25.8%, indicating that better-performing models are less susceptible to interference but still more so than the effect of prompt modification.63% similarUnverifiedAblation analysis shows that fine-tuning on the segment paradigm yields a +5.29-point improvement on the test set, while replacing the first 4,500 transcript words with 2,000 retrieved words yields only a +1.55-point improvement.63% similarUnverifiedOf the 11 LoRAs, only 3 produced meaningful effects, and 4 scored lower than the no-LoRA control group62% similarUnverifiedOn the frozen test set, the delivered system (LoRA + rules) caught 73 (recall 0.86), with 11 confident errors, 1 "not sure", and 0 false positives, outperforming the keyword baseline, LoRA-only fine-tuning, and zero-shot Qwen3.5-4B.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1114024API
curl https://kongchang.com/api/v1/knowledge/claims/1114024MCP
get_claim(id=1114024)