Unverified50% confidenceBenchmarkExact time

Qwen3-1.7B's faithfulness is 68% on 2-hop tasks but plummets to 30% on 6-hop tasks, whereas Qwen3-4B remains relatively stable across different hop counts.

1

Sources

50%

Confidence

Long-term

Relevance

9/24/2026

First Seen

Sources

Is Chain-of-Thought Actually Reasoning? Causal Measurements Reveal the Truth About CoT Faithfulness

Related Entities

Qwen3-1.7Bmodel Qwen3-4Bmodel

Related Claims

UnverifiedQwen3-1.7B's overall causal faithfulness is only 54.8%, far below that of Qwen3-4B.78% similar VerifiedQwen3.5's parameter activation rate is 4.3%, while DeepSeek V3's 671B model has an activation rate of 5.5%.75% similar UnverifiedIn the Terminal Bench test, Qwen 3.6 27B scored approximately 60.7, while Muse Glimmer scored only 51.7, falling significantly behind.74% similar VerifiedQwen3.6 27B scored 77.2 on SWE-bench Verified.74% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/946177

API

curl https://kongchang.com/api/v1/knowledge/claims/946177

MCP

get_claim(id=946177)