The　multi-agent　argument-based　negotiation　is　used　to　resolve　the　conflicts　in　supply　chain　collaboration,　and　to　change　the　preferences　of　negotiators　in　dynamic　environment　in　order　to　enhance　the　agent＇s　intelligence　and　adaptivity.　We　learn　opponents＇　preferences　by　PBIL　and　modify　PBIL＇s　learning　rate　by　adding　environmental　factor　to　improve　PBIL＇s　adaptivity.　We　verify　the　effectiveness　of　PBIL　in　negotiation　and　compare　PBIL　with　adaptive　PBIL　by　an　example.　The　results　show　that　PBIL　could　effectively　resolve　the　conflicts　in　supply　chain　collaboration,　adaptive　PBIL　could　get　a　stable　solution　in　dynamic　environment　and　promote　the　supply　chain　collaboration.　©　2014　IEEE.