All changes can be found here:
https://github.com/rbchan/unmarked/commits/master.
The following is just a summary.

Version 1.1.0
  o Add nmixTTD fitting function
  o Add experimental random effects support and TMB engine to occu and pcount
  o Add openMP support to some fitting functions (occuRN, gdistsamp, gmultmix, gpcount) for calculating likelihood in parallel
  o Define STRICT_R_HEADERS in C++ code for compatibility with future Rcpp update
  o Many bugfixes mainly related to predict()

Version 1.0.1
  o Fix LTO mismatches
  o Automatically convert characters to factors in unmarkedFrames
  o Many bugfixes, mainly related to predict()

Version 1.0.0

  o New functions 'distsampOpen' and 'multmixOpen' - open population versions of distsamp/gdistsamp and multinomPois/gmultmix
  o Add 'predict' method for output from 'ranef', for generating posterior samples of the random effects and running a function on them
  o Predict now correctly handles formulas containing functions and newdata with invalid factor levels
  o Remove reshape2 dependency
  o Bugfixes

Version 0.13-1.9000

  o New 'makePiFun' function factories added to generate piFuns with required defaults

  o 'plot' method for parboot now only "ask"s for new page when screen device is in use; restores former "ask" setting on exit.

Version 0.13-1

  o Fixes for compatibility with R 4.0


Version 0.13-0

  o New 'occuMS' function added for fitting multi-state occupancy models (single-season and dynamic)
  o New 'occuTTD' function for fitting continuous time-to-detection occupancy models (single season and dynamic). Thanks to Jonathan Cohen for help with this
  o New 'crossVal' function for doing cross-validation on fitted unmarked models and fitLists
  o New 'vif' function for calculating variance inflation factors for fitted unmarked models
  o Add ability to use complimentary-log-log link function in occu
  o Add built-in dependent double observer pi function
  o New C++ engines for gmultmix, gdistsamp, multinomPois, occuRN
  o Approximate integrals in C++ engines with trapezoidal rule function instead of using Rdqags
  o Misc minor bugfixes


Version 0.12-3

  o New 'occuMulti' function added by Ken Kellner


Version 0.12-0

  o Fixed mistake in turnover calculations in colext vignette (thanks
    to Giacomo Tavecchia) 

  o added pcount.spHDS from AHM book. 
  
  o updated predict method for pcount to include ZIP model
  
  o Adam Smith added some parallel capabilities to the parboot functionality
  
  o Adam Smith fixed formatMult conversion of julian date to factor
  
  o Auriel Fournier fixed formatDistData to pad data with NA
  
  o fixed error in obsToY for custom pi function
  
  
Version 0.11-0

  o Andy Royle is the new maintainer  

  o Added Rebecca Hutchinson's penalized likelihood function occuPEN
    (experimental) 

  o fixed bug in gmultmix to accommodate mixed sampling protocols (NA
    in count frequency vector is not counted in the constraint that
    multinomial cell probabilities sum to 1)  

  o Changed variable 'ufp' to 'ufc' in ovenbird data and related functions.

  o Removed constraint in pcountOpen that lambdaformula==omegaformula 

  o Fixed bug in gdistsamp that caused error when NAs were present in
    half-normal model  

  o Fixed bug in ranef (it was giving an error message for pcountOpen with the 
    new dynamics options (Ricker and Gompertz) and working incorrectly for
    pcountOpen with immigration)

  o Fixed bug in pcountOpen that occurred when covariates were time
    varying but not varying among sites 


Version 0.10-6

  o Fixed bug in C++ code that was causing problems on Solaris 


Version 0.10-5

  o Added new models of population dynamics to pcountOpen. Most
    changes contributed by Jeff Hostetler.  


Version 0.10-4

  o Added importFrom("plyr", "ldply") to NAMESPACE because "reshape" no longer
    depends on "plyr"


Version 0.10-3

  o RcppArmadillo was moved from "Depends" section of DESCRIPTION file to
    "LinkingTo"


Version 0.10-2

  Bug fixes

  o Thanks for Dirk Eddelbuettel for patch to deal with change in Armadillo's
    in-place reshape function. Serious problems might occur if you use a recent
    version of RcppArmadillo and an old version of unmarked.

  o Dave Miller added another NA handling fix in occuFP(). I forgot to add
    this one in the previous version.


Version 0.10-1

  o Doc fixes requested by CRAN


Version 0.10-0

  Bug fixes

  o  Fixed NA handling in occuFP()

  o  Fixed integration setting in C++ code that were causing segfaults
     when calling distsamp

  o  Replace raster:::layerNames() with raster:::names()


  New features

  o  distsamp() and gdistsamp() should be faster and more stable for
     some keyfun/survey combinations



Version 0.12-0