Unverified50% confidenceBenchmarkExact time
In reinforcement learning post-training, Deletion Tuning's prior-task accuracy remains almost at the base model level, whereas full fine-tuning, LoRA, and random sparse fine-tuning lose up to 5.4 percentage points.
1
Sources
50%
Confidence
Long-term
Relevance
11/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedOn math tasks, as the training answers become increasingly absurd, the proportion of misaligned responses soars from 4% to over 60%, while the model's general capabilities show almost no loss.78% similarUnverifiedWhen using the multi-agent verification process, MARCH's accuracy was suppressed to 4.4%, whereas when the same model was directly asked which solution was better, accuracy reached 43.7%.75% similarUnverifiedOn the temporal split test set (878 comparisons from 655 trials), a gradient-boosted emulator trained on registry data eliminated about one-fifth of the error of a simple baseline.74% similarUnverifiedIn reduced experiments using only 20%, 40%, and 80% of the training data, the performance gains persist stably, confirming the reliability of this method in low-resource settings.73% similarVerifiedResearch indicates that at Q4 precision and above, performance loss is within 1–3% for most tasks.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1153750API
curl https://kongchang.com/api/v1/knowledge/claims/1153750MCP
get_claim(id=1153750)