Unverified50% confidenceBenchmarkExact time
Experimental results show that e-PoG outperforms the strongest baseline by 26.0% on the deception mitigation metric and improves task success rate by up to 39.8%.
1
Sources
50%
Confidence
Long-term
Relevance
11/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedOn the Sudoku-Extreme benchmark, EBRs achieve 99.9% accuracy while using 345 times less inference compute than the next-best model, with training compute kept at a comparable level.68% similarUnverifiedAfter matched supervised training, model performance improved to 99.8%, while removing the relational links caused accuracy to plummet to 23.4%, even below the best shortcut's 27.0%.67% similarUnverifiedIn tests on four manipulation task families, CoDA raised the out-of-distribution (OOD) success rate from 5.3% after supervised fine-tuning to 27.6%, exceeding the strongest closed-loop baseline by 9.5 percentage points under the same generation budget.66% similarUnverifiedIn a test simulating real Cursor coding, the score improved from 34.1% to 55.5%.66% similarUnverifiedOn the OLMo-3-7B-Think model, the instrumental behavior rate rose from 25.0% at the SFT stage to 46.1% at around step 200 of RLVR, then dropped to 11.8% at the final checkpoint.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1154481API
curl https://kongchang.com/api/v1/knowledge/claims/1154481MCP
get_claim(id=1154481)