Soil potassium (K) occurs in multiple fractions, but the effects of integrated nutrient management (INM) on K dynamics remain poorly characterised in calcareous alluvial soils of the middle Indo-Gangetic Plains (IGP). A field experiment was conducted in an ongoing rice–wheat cropping system (RWCS) at Dr Rajendra Prasad Central Agricultural University (RPCAU), Pusa, Bihar, to evaluate eight nutrient management treatments involving nitrogen, phosphorus and potassium (NPK) fertilisation, cereal residue recycling, green manuring and Pusa Decomposer (PD) application on soil K fractions and crop productivity. Treatments combining cereal residue and green manure with residue decomposer (T5) recorded the highest water-soluble potassium (WSK) (7.26 mg kg-1), exchangeable potassium (EK) (64.39 mg kg-1), 1 hr tetraphenylboron-extractable potassium (TBK) (2002.96 mg kg-1) and 1 hr non-exchangeable potassium extracted using sodium tetraphenylboron (NEKNaTPB) (1931.31 mg kg-1). Residue decomposer application enhanced labile and kinetically accessible K within corresponding treatment pairs. At the same time, 168 hr non-exchangeable potassium (NEK), lattice potassium (LK) and total potassium (TK) remained statistically unchanged, indicating limited alteration of stable K reserves over four years. Rice and wheat grain and straw yields were highest under treatments receiving greater proportions of inorganic N and P (T1, T6 and T7), whereas T4 and T5 produced lower yields despite higher labile K. Labile K fractions were positively inter-correlated, with the strongest association between 1 hr TBK and 1 hr NEK (r = 0.82, p ≤0.001), whereas soil K fractions showed weak and non-significant relationships with crop yield. However, nutrient management enhanced labile K pools without significantly altering stable K reserves or explaining yield variation. Recycling cereal residue with a decomposer, together with 25 % substitution of fertiliser N, may sustain rice–wheat productivity while improving labile K availability in calcareous soil.