The capability of Cosmic Inflation to explain the latest Cosmic Microwave Background and Baryonic Acoustic Oscillation data is assessed by performing Bayesian model comparison within the landscape of nearly three-hundred models of single-field slow-roll inflation. We present the first Bayesian data analysis based on the third-order slow-roll primordial power spectra. In particular, the fourth Hubble-flow function ε₄ remains unbounded while the third function verifies, at two-sigma, ε₃ ∈[-0.4,0.5], which is perfectly compatible with the slow-roll predictions for the running of the spectral index. We also observe some residual excess of B-modes within the BICEP/Keck data favoring, at a non-statistically significant level, non-vanishing primordial tensor modes: log(ε₁) > -3.9, at 68% confidence level. Then, for 287 models of single-field inflation, we compute the Bayesian evidence, the Bayesian dimensionality and the marginalized posteriors of all the models' parameters, including the ones associated with the reheating era. The average information gain on the reheating parameter R _reh reaches 1.3 ± 0.18 bits, which is more than a factor two improvement compared to the first Planck data release. As such, inflationary model predictions cannot meet data accuracy without specifying, or marginalizing over, the reheating kinematics. We also find that more than 40% of the scenarios are now strongly disfavored, which shows that the constraining power of cosmological data is winning against the increase of the number of proposed models. In addition, about 20% of all models have evidences within the most probable region and are all favored according to the Jeffreys' scale of Bayesian evidences.

中文翻译：

---

宇宙膨胀处于十字路口


通过在近三百个单场慢滚暴胀模型的范围内进行贝叶斯模型比较，评估了宇宙暴胀解释最新宇宙微波背景和重子声振荡数据的能力。我们提出了第一个基于三阶慢滚原始功率谱的贝叶斯数据分析。特别是，第四个哈勃流函数 ε ₄保持无界，而第三个函数在 2-sigma 处验证 ε ₃ ∈[-0.4,0.5]，这与光谱运行的慢滚预测完全兼容指数。我们还观察到一些残留过量乙-BICEP/Keck 数据中的模式在非统计显着水平上有利于非消失的原始张量模式：log(ε ₁ ) > -3.9，置信水平为 68%。然后，对于 287 个单场暴胀模型，我们计算了所有模型参数（包括与再加热时代相关的参数）的贝叶斯证据、贝叶斯维数和边缘化后验。再加热参数的平均信息增益右_reh达到 1.3 ± 0.18 位，与第一个普朗克数据发布相比，提高了两倍多。因此，如果不指定或边缘化再加热运动学，通货膨胀模型预测就无法满足数据准确性。我们还发现，超过 40% 的场景现在被强烈反对，这表明宇宙学数据的约束力正在战胜提出模型数量的增加。 此外，大约 20% 的模型在最可能的区域内有证据，并且根据贝叶斯证据的杰弗里斯尺度，所有模型都受到青睐。