The data set is a collection of 257 unique data points collected from different sources literature around the world. Each Data point is composed of proximate, ultimate, structural properties and the High Heating Value(HHV) of biomass (agricultural and municipal solid organic waste). Proximate properties include the Ash content(Ash), Volatile Matter(VM), and Fixed Carbon(FC) of the biomass. Ultimate properties include the elemental Carbon(C), Hydrogen(H), Oxygen(O) and Sulphur(S). Structural/lignocellulosic composition including Cellulose(Cel), Hemicellulose(Hemi) and Lignin(Lig) is also included together with the HHV(MJ/kg). All the compositional columns are presented as percentage (wt%) on dry basis while HHV is presented in MJ/kg. The column “Biomass” names the material as described in the source (for example, rice husk, sugarcane bagasse, paper sludge, mixed food waste). The column “Source” provides the bibliographic reference or identifier from which the observation was extracted. Random missing values were imputed with k-nearest-neighbours to yield a complete, analysis-ready table; no generated (cGAN-produced) samples are stored here. The accompanying manuscript tests the hypothesis that lignocellulosic composition and heating value can be predicted from standard fuel analyses even with a small, heterogeneous data set and that accuracy improves when conditional generative adversarial networks are used to increase the training distribution and when multi-output ensemble regressors using chained targets are used to capitalize on cross-property correlations. Users need to read totals accordingly: ash + volatile matter + fixed carbon is approximately 100 wt% (dry), elemental totals are slightly less than 100 wt% due to occasional by-difference oxygen, and cellulose + hemicellulose + lignin do not necessarily sum to 100 wt% because extractives or minor constituents are sometimes reported separately by some sources. The descriptive statistics cited in our article refer to a winsorised copy used for robustness checks; the file deposited here is the harmonized, literature-derived dataset.