/reasoning-multi-path

tools:
  - read
  - write
  - grep
  - bash
arguments: $PROBLEM

Multi-Path Reasoning

Execute parallel reasoning exploration across multiple solution paths to find optimal approaches.

Based on the WFGY project: https://github.com/onestardao/WFGY

Instructions

1. **Initialize Multi-Path Setup**

• Parse problem from "$PROBLEM"
• Set number of parallel paths (default: 5)
• Load current context and constraints
• Define exploration parameters:
• Divergence factor: 0.3
• Pruning threshold: 0.1
• Max iterations: 10

2. **Generate Reasoning Paths**

• For each path i (1 to N):
• Apply unique perturbation V_i
• Vary approach angle:
• Path 1: Direct/logical
• Path 2: Creative/lateral
• Path 3: Systematic/methodical
• Path 4: Analogical/comparative
• Path 5: Contrarian/inverse
• Maintain problem constraints
• Track path evolution

3. **Evolve Paths Iteratively**

• For each iteration:
• Advance each path independently
• Calculate path metrics:
• Progress score
• Confidence level
• Semantic coherence
• Logic consistency
• Apply selection pressure:
• Amplify successful paths
• Diminish failing paths
• Cross-pollinate insights

4. **Weight and Rank Paths**

• Calculate path probabilities:
• P_i = exp(-ΔS_i) / Σexp(-ΔS_j)
• Compute dynamic weights:
• W_i based on progress rate
• Adjust for confidence
• Rank by composite score:
• Score = P_i * W_i * Confidence_i

5. **Synthesize Solutions**

• Combine top paths weighted by score
• Extract unique insights from each
• Identify convergent conclusions
• Note divergent possibilities
• Create unified solution

Output Format

MULTI-PATH REASONING ANALYSIS
═══════════════════════════════════════
Problem: "$PROBLEM"
Paths Explored: 5
Iterations: [count]

Path Evolution:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
        Iteration
Path    1  2  3  4  5  6  7  8  9  10
────────────────────────────────────────
Path 1  ▪──●──●──●──●──●──●──●──●──●  85%
Path 2  ▪──○──●──●──●──●──●──●──○──○  72%
Path 3  ▪──○──○──●──●──●──○──○──×──×  Failed
Path 4  ▪──○──○──○──●──●──●──●──●──●  78%
Path 5  ▪──●──●──●──●──●──●──●──●──●  91%

Legend: ▪ Start ○ Exploring ● Promising × Failed

Top 3 Solution Paths:
────────────────────────────────────────

PATH 5 (Best) - Score: 0.91
Approach: [Contrarian/Inverse thinking]
Key Insight: [Main discovery]
Reasoning: [Brief explanation]
Confidence: 91%

PATH 1 - Score: 0.85
Approach: [Direct logical analysis]
Key Insight: [Main discovery]
Reasoning: [Brief explanation]
Confidence: 85%

PATH 4 - Score: 0.78
Approach: [Analogical comparison]
Key Insight: [Main discovery]
Reasoning: [Brief explanation]
Confidence: 78%

Convergent Findings:
• All paths agree: [Common conclusion 1]
• 4/5 paths suggest: [Common conclusion 2]
• Majority indicates: [Common conclusion 3]

Divergent Possibilities:
• Path 2 uniquely suggests: [Alternative]
• Path 5 contrarian view: [Opposite angle]

Synthesized Solution:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[Weighted combination of successful paths]
[Incorporating insights from all approaches]
[Final recommended solution]

Confidence: [weighted average]%
Risk Assessment: [Low/Medium/High]

Alternative Approaches:
1. If [condition], consider Path 2
2. If [constraint], prefer Path 4
3. For [scenario], Path 1 is optimal

Path Statistics:
- Average Convergence: [value]
- Cross-Path Correlation: [value]
- Solution Stability: [value]

Path Types

Logical Path

• Step-by-step deduction
• Rule-based reasoning
• Systematic progression

Creative Path

• Lateral thinking
• Unexpected connections
• Novel approaches

Empirical Path

• Evidence-based
• Data-driven
• Experimental validation

Analogical Path

• Pattern matching
• Similar problem spaces
• Transfer learning

Contrarian Path

• Inverse thinking
• Challenge assumptions
• Devil's advocate

Configuration

{
  "num_paths": 5,
  "max_iterations": 10,
  "divergence_factor": 0.3,
  "pruning_threshold": 0.1,
  "selection_pressure": 0.2,
  "crossover_rate": 0.15
}

Advanced Usage

# Explore more paths
/reasoning:multi-path "problem" --paths 10

# Focus on creative solutions
/reasoning:multi-path "problem" --emphasize creative

# Quick exploration
/reasoning:multi-path "problem" --iterations 5

# Deep exploration
/reasoning:multi-path "problem" --iterations 20

Integration

Multi-path reasoning works with:

• `/wfgy:bbpf` for path generation
• `/semantic:node-build` to record paths
• `/boundary:detect` to check path safety
• `/reasoning:chain-validate` to verify logic