The paper introduces the "anti-$k_t$" jet clustering algorithm, a new member of the class of sequential recombination jet algorithms. Unlike traditional algorithms like $k_t$ and Cambridge/Aachen, which are parametrized by the power of the energy scale in the distance measure ($p = 1$ for $k_t$ and $p = 0$ for Cambridge/Aachen), the anti-$k_t$ algorithm uses a negative value of $p$. This results in a behavior similar to an idealized cone algorithm, where jets with only soft fragmentation are conical, and the jet boundaries are resilient to soft radiation but flexible to hard radiation. Key properties of the anti-$k_t$ algorithm include: - **Jet Shape**: Soft particles cluster with hard particles before clustering among themselves, leading to conical jets. - **Area Properties**: The passive and active areas of jets are independent of the soft particle configuration, and the anomalous dimensions are zero. - **Back Reaction**: The probability of back-reaction is suppressed by the jet transverse momentum, reducing the impact of underlying event (UE) and pileup contamination. - **Non-Global Logarithms**: The single-logarithmic non-global terms are identical to those of ideal cones, simplifying their determination. - **Milan Factor**: The Milan factor retains its universal value, indicating that the jet boundary is not affected by soft radiation. - **Computational Efficiency**: The algorithm is as fast as $k_t$ clustering for $N \lesssim 20000$ particles. The paper also discusses the practical implications of these properties, including their application in top mass reconstruction in LHC $t\bar{t}$ events. The anti-$k_t$ algorithm is found to perform well in this context, suggesting it could be a good replacement for iterative cone algorithms used in experiments like CMS.The paper introduces the "anti-$k_t$" jet clustering algorithm, a new member of the class of sequential recombination jet algorithms. Unlike traditional algorithms like $k_t$ and Cambridge/Aachen, which are parametrized by the power of the energy scale in the distance measure ($p = 1$ for $k_t$ and $p = 0$ for Cambridge/Aachen), the anti-$k_t$ algorithm uses a negative value of $p$. This results in a behavior similar to an idealized cone algorithm, where jets with only soft fragmentation are conical, and the jet boundaries are resilient to soft radiation but flexible to hard radiation. Key properties of the anti-$k_t$ algorithm include: - **Jet Shape**: Soft particles cluster with hard particles before clustering among themselves, leading to conical jets. - **Area Properties**: The passive and active areas of jets are independent of the soft particle configuration, and the anomalous dimensions are zero. - **Back Reaction**: The probability of back-reaction is suppressed by the jet transverse momentum, reducing the impact of underlying event (UE) and pileup contamination. - **Non-Global Logarithms**: The single-logarithmic non-global terms are identical to those of ideal cones, simplifying their determination. - **Milan Factor**: The Milan factor retains its universal value, indicating that the jet boundary is not affected by soft radiation. - **Computational Efficiency**: The algorithm is as fast as $k_t$ clustering for $N \lesssim 20000$ particles. The paper also discusses the practical implications of these properties, including their application in top mass reconstruction in LHC $t\bar{t}$ events. The anti-$k_t$ algorithm is found to perform well in this context, suggesting it could be a good replacement for iterative cone algorithms used in experiments like CMS.