Diffusion imaging (DI) at sub-millimeter isotropic resolution is challenging due to small voxel size and low SNR. Slider-SMS acquisition was recently proposed (1) to provide a large SNR efficiency gain by acquiring a large number of imaging slices simultaneously to increase volume encoding. This approach combines Blipped-CAIPI SMS parallel imaging (2) with slice super-resolution (Slider) acquisition (3), where e.g. 6 slices can be acquired simultaneously through MB-2 and 3x-Slider. Here, the Slider approach relies on multiple thick-slice acquisitions with sub-voxel shifts to enable a super-resolution reconstruction. Since the shifted thick-slices do not form an orthogonal encoding basis, regularization is required to improve the conditioning of the reconstruction. With Tikhonov regularization, high quality 3x-Slider reconstruction has been achieved with ~√3 SNR benefit at a cost of some spatial blurring (~25-30% side-lobes from adjacent voxels). However, higher Slider factors have not been possible without large blurring. In this work, we overcome this limitation by developing generalized Slider (gSlider), which exploits RF encoding to improves the orthogonality of Slider’s encoding basis. We show that gSlider can be used to acquire 5 slices simultaneously to provide close to √5 SNR gain, while retaining sharp slice/partition resolution, comparable to that of conventional 2-D slice-selective imaging. Through a combined gSlider-SMS acquisition (5x-gSlider and MB-2), we demonstrate a highly efficient 10 simultaneous slice acquisition for high quality whole-brain 660μm isotropic DI. Through the use of appropriately designed RF pulses, magnitude and/or phase profiles of thick-slice acquisitions in gSlider can be tailored to minimize their linear dependencies and improve the conditioning of the super-resolution reconstruction. A key feature is that the thick-slice profiles acquired form highly independent basis, while maintaining high SNR in each individual thick-slice acquisition. The high SNR at acquisition allows for proper estimation and removal of the DI’s phase corruption (without lengthy navigators), and thus the use of real-valued rather than magnitude data to avoid large signal bias that leads to poor super-resolution reconstruction and diffusion processing. Fig1 shows an exemplar basis for 5x-gSlider, where DIST RF pulses (4) are used to create high quality 90° excitations with ‘slice-phase dithering’ (in conjunction with conventional 180° SLR refocusing). Shown are the RF pulses and corresponding excitation profiles, where for each excitation, one sub-slice undergoes a π phase modulation. This ‘slice-phase dithering’ approach provides a highly independent basis while maintaining high SNR for each thick-slice, at ~3x that of the thin-slice. Using the impulse response and SNR characterization of Tikhonov super-resolution reconstruction outlined in Fig2, this 5x-gSlider can achieve an SNR gain of √4.6, while maintaining sharp partition resolution at 7.5% side-lobes amplitude (Fig1-right), comparable to conventional 2D imaging. gSlider-SMS data were acquired in a volunteer on the 3T CONNECTOM system using custom-built 64-channel array. Fig3 illustrates the 10 simultaneous slice sagittal acquisition (gSlider×MB = 5×2), where ZOOPPA (5) is also employed with outer volume suppression of the neck and phase-encoding in head-foot to provide low distortion with whole-brain imaging capabilities (R_zoom×R_GRAPPA = 1.85×2). Imaging parameters were: 660μm iso; FOV 220×118×151.8 mm³; p.f. 6/8; TE/TReff = 80ms/22s (TR per thick-slice volume = 4.4s); effective echo spacing = 0.32ms, 4 repetitions of 64 directions at b=1500 s/mm² with interspersed b0 every 10 volumes, total scan-time ~100 min. Background phase removal was performed using the real value diffusion algorithm (6), followed by Tikhonov regularized super-resolution reconstruction. Eddy-current correction and DTI-fit were accomplished using FSL (7). Fig4 demonstrates diffusion reconstruction of gSlider-SMS, where 660μm isotropic resolution enables detection of fine-scale structures in any spatial orientation; with multiple voxels across cortical depth and expected dark bands of FA at gray-white tissue boundaries. Fig5 compares reconstructions from 1, 2 and 4 repetitions of 64-direction dataset (25min/repetition). In moving from one to two repetitions, the FA map improves significantly with lower noise level, providing the ability to obtain reasonable tensor results at this extreme resolution in 50 minutes. With 4 repetitions, the FA and tensor estimates further improve to provide robust diffusion estimates. We proposed a novel basis encoding super-resolution method and demonstrated its ability for efficient DI acquisition with 10 simultaneous slices to provide > 3-fold gain in SNR efficiency. An exemplary slice-phase dithering method was developed to provide near orthogonal slice encoding basis while maintaining high SNR for individual thick-slice acquisition to enables background phase removal without additional lengthy phase navigators. Future work will further explore other encoding basis under the gSlider framework such as a combined slice-shifting and phase-dithering approach and thick-slice Hadamand encoding with odd total number of sub-slices which does not result in near complete signal cancellation.