Tech Meridian ← ENTITY INDEX
PROMY MERIDIAN RU

COMPANY · ENTITY #7736

Gemma4

Related event timeline, sources and context from the news index.

EVENT TIMELINE

2

RESEARCH · 1 SOURCE · arXiv cs.AI

Answer configuration determines whether LLMs follow their reasoning trace or current evidence

The arXiv preprint (2610.20805v1) studies how keeping a model's reasoning trace at answer time interacts with verified evidence by comparing two answer configurations: continuation of the model's own thinking versus a separate answer call. Across gpt-oss, Gemma4 and Qwen3 on HotpotQA and MuSiQue, using a separate answer call raises evidence-following by 29.9–60.6 points versus continuation; removing the trace lowers accuracy similarly under both configurations, but continuation disproportionately repeats stale conclusions while a separate call follows current evidence far more often. The paper also finds that the final line of a trace carries most of its value, gold evidence helps most before the trace is written, distractors weaken correction under a separate call, and renewed reasoning (as opposed to a verification instruction) typically follows current evidence in conflicts, leading the authors to recommend reporting both trace benefit and evidence-following alongside answer configuration when evaluating traces.

7.0

RESEARCH · 1 SOURCE · arXiv cs.AI

Policy-as-Skill: governed LLM decision-support runtime (arXiv:2609.27087v1)

The paper introduces Policy-as-Skill (PaS), a modular runtime that packages evidence validation, review routing, version control, and auditability as executable, versioned policy capabilities. Evaluated with 13 methods on 600 development tasks using a fixed Gemma4 backend, PaS+Audit reported 53.8% exact accuracy, macro-F1 0.346, review F1 0.854, citation precision 1.000, policy-reference recall 0.984, and audit completeness 1.000, and the authors report that deterministic control can raise aggregate accuracy to 61.2% though results are task-dependent; PaS+Audit outperforms LLM+RAG on most governance and review metrics.

6.0