cd /news/large-language-models/from-imperfective-paradox-to-moorean… · home › topics › large-language-models › article
[ARTICLE · art-171563] src=machinebrief.com ↗ pub= topic=large-language-models verified=true sentiment=· neutral

From Imperfective Paradox to Moorean Absurdity: Single-Chain Reasoning Lock-In and Multi-Candidate Verbalization Across Aspectual and Pragmatic Ambiguity

A preregistered study of Qwen3-8B-Q8_0 across 16,640 formal units at temperature=0.0 found that prompt-and-readout protocols shift the model's judgments on aspectual and pragmatic ambiguity, with Direct-CoT and Multi-Candidate Verbalization (MCV) differing by 25.00 percentage points on telic high-confidence completion (Holm p=1.31239e-05) while NLI-CoT collapsed by -91.25 pp on uninterrupted atelic items (Holm p=1.81805e-21). In Experiment 4, MCU-CF raised perceived coherence on implicit Moorean sentences by 26.64 pp over Direct-CoT (Holm p=0.0006), alongside Direct-CF minus Direct-CoT at 29.15 pp and MCU-CoT minus Direct-CoT at 17.31 pp. The authors report 15 of 16 hypothesis slots retained confirmatory status (12 rejecting H0 under within-experiment Holm control at alpha_fam=0.05, with E3_H3 demoted to descriptive), and a 1,800-call OpenAI ablation plus a stage="dev" P16.7 cross-model extension (4,222 valid of 4,240 planned/collected units across five external models) showed discrete NLI-CoT and scalar Direct-CoT judgments dissociate across models; the authors state internal premature lock-in remains a hypothesis, and code, stimuli and analysis plans are at github.com/asj252/llm-judgment-studies.

recommendation

Test both discrete NLI-CoT and scalar Direct-CoT readouts before trusting a model's entailment judgments; the source measured a -91.25 pp collapse under NLI-CoT on atelic items.

more in playbook: Evals and testing for LLM apps
by read1 min views1 publishedNov 18, 2026

arXiv:2611.22188v1 Announce Type: new Abstract: Natural language entailment and coherence judgments often depend on whether an event merely occurred, reached a telic culmination, satisfied a plan, or was evaluated under multiple pragmatic standards. We test observable consequences of the single-chain premature lock-in hypothesis across four preregistered, sealed formal experiments on Qwen3-8B-Q8_0 (16,640 formal units, temperature=0.0): 15 of 16 hypothesis slots retain confirmatory status (12 rejecting H0 under within-experiment Holm control at alpha_fam=0.05; E3_H3 demoted to descriptive status). In Experiment 1, Direct-CoT vs. MCU differs by 25.00 pp on telic high-confidence completion (Holm p=1.31239e-05), while NLI-CoT collapses by -91.25 pp on uninterrupted atelic items (Holm p=1.81805e-21). Multi-Candidate Verbalization (MCV) elicits K=5 weighted estimates in one call, separating the prompt from unweighted (MCU) and self-weighted (MCW) readouts. Mandarin (E2) and English (E3) expose an Interruption-Anchored Realis Paradox under NLI-CoT. In Experiment 4, MCU-CF raises perceived coherence on implicit Moorean sentences by 26.64 pp over Direct-CoT (Holm p=0.0006), alongside Direct-CF - Direct-CoT = 29.15 pp and MCU-CoT - Direct-CoT = 17.31 pp. A 1,800-call OpenAI ablation and a stage="dev" P16.7 cross-model extension (4,240 planned/collected units across five external models: 4,222 valid, 18 format-missing, 0 transport-missing; <2% missing per slice; OpenAI gpt-4o-mini at top_logprobs=20 achieves 1,000/1,000 valid units) show that discrete NLI-CoT and scalar Direct-CoT judgments dissociate across models. Confirmatory effects concern observable prompt-and-readout protocols; internal premature lock-in remains a hypothesis. All code, stimuli, preregistered analysis plans, and verification tools are available at https://github.com/asj252/llm-judgment-studies.

── more in #large-language-models 4 stories · sorted by recency
── more on @qwen3-8b-q8_0 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/from-imperfective-pa…] indexed:0 read:1min 2026-11-18 · —