Unverified50% confidenceBenchmarkExact time
On Autonomous Science Bench, SciMARS + Memory solved 2 of 18 GLM-5.3 attempts, while the base multi-agent team solved only 1/18.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
11/10/2026
First Seen
Valid until: 11/24/2026
Sources
Related Entities
Related Claims
UnverifiedIn a test of 100 tasks randomly sampled from an internal binary exploitation benchmark by Anthropic's red team, GLM-5.3 successfully developed a complete control flow hijack in 4% of trials.66% similarUnverifiedIn a preregistered study, confidence intervals relying solely on LLM judges failed to cover the full human-annotation ground truth in 18 of 21 task-and-judge combinations.62% similarUnverifiedThe two sets of experiments on BraTS with 20% and 30% axial slice retention report the best DSC/HD95 obtained after retraining each model separately, rather than directly testing the same weights across degradation levels.61% similarUnverifiedThe researchers found that when χ1 is close to 1, the test error decreases monotonically in all four experimental units.61% similarVerifiedExperiments by the Long-Term team showed that using the same model with only adjustments to rule constraints and the memory system raised the score from 52.8% to 66.5%, jumping the ranking from Top 30 to Top 5.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1154741API
curl https://kongchang.com/api/v1/knowledge/claims/1154741MCP
get_claim(id=1154741)