Topologically massive gravity formulated in terms of $\Gamma$ or $\omega$ is the same; for example in Maloney's paper on the geometric microstates of the three-dimensional black hole, he uses the $\Gamma$ formulation, and Witten's talk on revisiting $2+1$-dimensional gravity includes the Chern-Simons term with the spin connection $\omega$, but they are both describing the same theory.
One subtlety I'd like to point out, when performing reformulations of fields, is it can modify it in a non-trivial way. In particular, gravity in $2+1$ dimensions is considered trivial, but when formulated in terms of a gauge field out of the connection and vierbein,
$$A = \begin{pmatrix}
\omega & e\\
-e & 0
\end{pmatrix}$$
allowing a non-invertible $e$ and thus non-classical configurations. It turns out this makes the Einstein-Hilbert action ill-defined unless the coupling is quantised.
Since you seem interested in topological massive gravity, and since it wasn't mentioned in the papers you listed, I would like to point out it has been conjectured by Witten and others that if the Frenkel-Lepowsky-Meurman conjecture is true, it may be that the dual $k=1$ CFT is the theory whose symmetry is described by the Monster group - I don't know if this avenue is of interest to you.
To explicitly show the equivalence, it seems rather messy. As a starting point, note the relation between the spin connection $\omega$ and the affine connection $\Gamma$ as,
$$\Gamma^\lambda_{\mu\nu} = e^\lambda_A \partial_\mu e^A_\nu + e^\lambda_A e^B_\nu {\omega_\mu}^A_B$$
where capital Roman indices denote the orthonormal basis and Greek indices the coordinate basis. We can substitute this into the Chern-Simons term for $\Gamma$, which is explicitly in index notation,
$$\mathcal L \sim \epsilon^{\lambda\mu\nu} \Gamma^\rho_{\lambda \sigma} \left( \partial_\mu \Gamma^\sigma_{\rho\nu} + \frac{2}{3} \Gamma^\sigma_{\mu\kappa} \Gamma^\kappa_{\nu\rho}\right),$$
using the definition of the wedge product and exterior derivative. Using the relation between the connections, we thus have,
$$\mathcal L \sim \epsilon^{\lambda\mu\sigma}\left(e^\rho_A \partial_\lambda e^A_\nu + e^\rho_A e^B_\sigma {\omega_\lambda}^A_B \right) \bigg[ \partial_\mu \left( e^\sigma_A \partial_\rho e^A_\nu + e^\sigma_A e^B_\nu {\omega_\rho}^A_B\right)+ \bigg.\\ + \bigg. \frac{2}{3} \left(e^\sigma_A \partial_\mu e^A_\kappa + e^\sigma_A e^B_\kappa {\omega_\mu}^A_B \right) \left( e^\kappa_A \partial_\nu e^A_\rho + e^\kappa_A e^B_\rho {\omega_\nu}^A_B\right)\bigg].$$
At this point it is a matter of tedious manipulation, which should hopefully show an equivalence to the theory in terms of the spin connection.