Geometry of Divergence: Tracking Hidden-State Trajectories for Adaptive Multi-Turn Reasoning
Not assessedPlan blockedFindingclaim-variance-slope-separability
Correct episodes exhibit higher variance slopes beta_t than incorrect episodes across all six task-LLM pairings. Separability is statistically significant (p < .001) for Math (Llama-3.1-8B), Code (Qwen3-14B), Retail (Qwen3-14B), and Retail (Qwen3-32B), but not statistically significant for Airline (Qwen3-14B, p=0.30; Qwen3-32B, p=0.057).
Source: paper:Page 4, Section 4.3, Paragraph 'Correct episodes exhibit higher variance slopes than incorrect episodes' and Figure 2(b)
Reported and observed measurements
mann_whitney_u_p_value
meas-pvalue-variance-slope-airline-14b
Reported 0.3 scalar
Observed — scalar
mann_whitney_u_p_value
meas-pvalue-variance-slope-airline-32b
Reported 0.057 scalar
Observed — scalar
Assessments (0)
No immutable Assessment has been published for this Claim yet.
Runs (0)
No execution has been linked to this Claim yet.