Environmental enrichment can have complex, contradictory effects on aggression and animal welfare. Although increasing enrichment may reduce aggression by limiting the visibility of competitors, it may also intensify territorial defense and harm the welfare of subordinates who are the target of territorial aggression. However, it remains unclear how variation in the complexity of an enrichment structure defended by a single dominant individual influences welfare outcomes for subordinates. In the cichlid fish Astatotilapia burtoni, dominant males defend structures as mating territories, whereas subordinate males do not. To establish social hierarchies, we housed two differently sized males (larger become dominant) and six females in one compartment containing one defendable structure. We manipulated structure complexity by placing 1, 2, or 3 halves of terracotta pots clumped together. Increasing cave number did not enhance territoriality: dominant males showed similar aggression rates, relative gonad sizes, and testosterone levels across treatments. Cave number did not significantly affect mortality, body condition, growth rate, or fin damage in subordinate males. Although relative gonad size and testosterone levels were generally higher in dominants, these differences were not always significant in the 2 and/or 3-cave treatments, suggesting weaker physiological differentiation between social states. The behavioral data supported this pattern, with status-specific differences in chase rates declining with increasing cave number. Overall, increasing environmental enrichment had little effect on welfare, but resulted in reduced physiological and behavioral distinctions between dominant and subordinate males. Therefore, careful consideration of enrichment strategies is essential for accurately interpreting status-specific outcomes in laboratory settings.
Surrogate Neural Architecture Codesign Package (SNAC-Pack)
arXiv:2512.15998v1 Announce Type: cross Abstract: Neural Architecture Search is a powerful approach for automating model design, but existing methods struggle to accurately optimize for real

