Unverified50% confidenceBenchmarkExact time
On the GenEval benchmark, LARC achieves a 7-percentage-point improvement over BAGEL, reaching state-of-the-art performance.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
11/2/2026
First Seen
Valid until: 11/16/2026
Sources
Related Entities
Related Claims
UnverifiedThe framework achieves an overall accuracy improvement of 6.9% over the current state of the art on the MERLIN benchmark.70% similarUnverifiedPrecision Mode leads the strongest frontier model chunk-and-merge baseline solution by 7 percentage points in accuracy.63% similarUnverifiedThree VLMs all improved in exact-match rate on four scientific chart benchmarks, with a single maximum improvement of 55.2 percentage points.61% similarUnverifiedOn 1,000 validation cases, the strongest LLM baseline achieves 62.7% accuracy, while EMBER (73.6%) and ORBIT (70.9%) perform better.60% similarUnverifiedTEMPEST outperforms the best classical model by 17.9 percentage points and the strongest triplet-loss baseline by 58.4 percentage points.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1125585API
curl https://kongchang.com/api/v1/knowledge/claims/1125585MCP
get_claim(id=1125585)