Diffusion Language Models: An Experimental Analysis
Diffusion Language Models: An Experimental Analysis
A new arXiv paper explores Diffusion Language Models (DLMs) as an alternative to autoregressive generation. It analyzes this emerging paradigm through experimental study.
Granularity-Regulated Adaptive Efficiency for Test-Time Scaling
A new arXiv paper proposes a granularity-regulated adaptive computational efficiency scheme for Test-Time Scaling (TTS). It optimizes the verifier to improve reasoning performance with better compute efficiency.
Uncertainty Decomposition for Clarification Seeking in LLM Agents
A new arXiv paper argues that classical uncertainty frameworks are insufficient for interactive LLM agents. It proposes a decomposed, communicable representation for uncertainty.
Which Pairs to Compare for LLM Post-Training?
A new arXiv paper analyzes data-collection strategies for preference-based post-training. It investigates how to select the best completion pairs for prompts to optimize alignment.
Hidden Anchors in Multi-Agent LLM Deliberation
A new arXiv paper investigates “Hidden Anchors” in multi-agent LLM deliberation. It models how agents exchange answers and identifies hidden biases similar to human decision-making.
Characterizing Narrative Content in Web-scale LLM Pretraining Data
A new arXiv paper presents the first fine-grained study of narrative features in the Dolma 3-trillion-token open corpus. It characterizes narrative content in web-scale pretraining data.
Claude Code v2.1.185: API Timeout Hint Extended to 20s
Claude Code v2.1.185 is released. It updates the API stall hint and extends the silence trigger timeout from 10s to 20s.
OpenClaw Releases 2026.6.20-alpha.1
OpenClaw releases version 2026.6.20-alpha.1. This is an alpha release.