[6/20 05:00] FP4 pretraining shrinkage bias / UFP4 / LLM psychological profiles measurement artifact

[6/20 08:00] Thermodynamic Measure of Intelligence / Calibrated MoE Under Distribution Shift

AI News Flash for June 20, 2026 at 08:00 UTC. Topics: 1. Thermodynamic Measure of Intelligence https://arxiv.org/abs/2606.20231 2. Calibrated MoE Under Distribution Shift https://arxiv.org/abs/2606.20544 3. Augmenting Game AI with Deep Reinforcement Learning https://arxiv.org/abs/2606.20210

[6/20 07:00] Multi-LCB multilingual code-generation benchmark / MACR knowledge-conflict resolution

AI News Flash for June 20, 2026 at 07:00 UTC. Topics: 1. Multi-LCB multilingual code-generation benchmark https://arxiv.org/abs/2606.20517 2. MACR knowledge-conflict resolution https://arxiv.org/abs/2606.20245 3. Lagrange end-to-end autonomous driving framework https://arxiv.org/abs/2606.20274

[6/20 06:00] ENPIRE robot policy self-improvement / Safety-aligned models and mixed compliance

AI News Flash for June 20, 2026 at 06:00 UTC. Topics: 1. ENPIRE robot policy self-improvement https://arxiv.org/abs/2606.19980 2. Safety-aligned models and mixed compliance https://arxiv.org/abs/2606.20508 3. SEVRA budget-aware selective verification https://arxiv.org/abs/2606.19808

[6/20 05:00] FP4 pretraining shrinkage bias / UFP4 / LLM psychological profiles measurement artifact

AI News Flash for June 20, 2026 at 05:00 UTC. Topics: 1. FP4 pretraining shrinkage bias / UFP4 https://arxiv.org/abs/2606.20381 2. LLM psychological profiles measurement artifact https://arxiv.org/abs/2606.20205 3. Process-verified RL for theorem proving via Lean https://arxiv.org/abs/2606.20068

[6/20 04:00] ITNet - learnable integral transform unifying convolution, attention, recurrence / Experimental analysis of diffusion language models

AI News Flash for June 20, 2026 at 04:00 UTC. Topics: 1. ITNet - learnable integral transform unifying convolution, attention, recurrence https://arxiv.org/abs/2606.19538 2. Experimental analysis of diffusion language models https://arxiv.org/abs/2606.19475 3. AgenticRei - deontic policies for runtime governance of agents https://arxiv.org/abs/2606.19464