File size: 37,422 Bytes
b25b8f2 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 | """
Streamlit Dashboard for Mathematical Reasoning Verification System
Interactive UI with real-time processing logs and results display
"""
import streamlit as st
import time
from typing import List, Dict, Any
from core import run_verification_parallel
# Initialize session state
if 'steps_log' not in st.session_state:
st.session_state.steps_log = []
if 'results' not in st.session_state:
st.session_state.results = None
def add_log(step: str, model: str, status: str, details: str):
"""Add entry to processing log."""
log_entry = {
"step": step,
"model": model,
"status": status,
"details": details,
"timestamp": time.time()
}
st.session_state.steps_log.append(log_entry)
def display_flowchart(problem="", steps_input=""):
"""Display interactive flowchart with expandable explanations."""
# Check if we have results to show status
has_results = st.session_state.results is not None
results = st.session_state.results if has_results else None
# Parse steps if provided
steps = []
if steps_input:
steps = [s.strip() for s in steps_input.split('\n') if s.strip()]
elif has_results:
steps = results.get('steps', [])
# Problem-Specific Flowchart
if problem or steps:
st.markdown("### 📊 Problem Flowchart")
st.markdown("**Problem:**")
st.info(problem if problem else "No problem entered yet")
if steps:
st.markdown("**Solution Flow:**")
# Create flowchart for the actual problem
flowchart_lines = []
flowchart_lines.append("```")
# Format problem text to fit in box
problem_display = problem[:45] + "..." if len(problem) > 45 else problem
problem_display = problem_display.ljust(50)
flowchart_lines.append("┌─────────────────────────────────────────────────────────┐")
flowchart_lines.append(f"│ 📥 PROBLEM: {problem_display} │")
flowchart_lines.append("└────────────────────┬────────────────────────────────────┘")
flowchart_lines.append(" │")
flowchart_lines.append(" ▼")
for i, step in enumerate(steps, 1):
# Extract key info from step
step_short = step[:45] + "..." if len(step) > 45 else step
# Check if this step has an error (if results available)
has_error = False
if has_results:
classified_errors = results.get('classified_errors', [])
for error in classified_errors:
if error.get('step_number') == i:
has_error = True
break
# Determine status icon
status_icon = "❌" if has_error else "✅"
# Format step text to fit in box (max 45 chars)
step_display = step_short.ljust(45)
if i < len(steps):
flowchart_lines.append(f"┌─────────────────────────────────────────────────────────┐")
flowchart_lines.append(f"│ {status_icon} STEP {i}: {step_display} │")
flowchart_lines.append("└────────────────────┬────────────────────────────────────┘")
flowchart_lines.append(" │")
flowchart_lines.append(" ▼")
else:
flowchart_lines.append(f"┌─────────────────────────────────────────────────────────┐")
flowchart_lines.append(f"│ {status_icon} STEP {i}: {step_display} │")
flowchart_lines.append("└────────────────────┬────────────────────────────────────┘")
flowchart_lines.append(" │")
flowchart_lines.append(" ▼")
flowchart_lines.append("┌─────────────────────────────────────────────────────────┐")
flowchart_lines.append("│ 📤 FINAL ANSWER │")
if has_results:
consensus = results.get('consensus', {})
final_verdict = consensus.get('final_verdict', 'UNKNOWN')
verdict_icon = "❌ ERROR" if final_verdict == "ERROR" else "✅ VALID"
verdict_display = verdict_icon.ljust(55)
flowchart_lines.append(f"│ {verdict_display} │")
flowchart_lines.append("└─────────────────────────────────────────────────────────┘")
flowchart_lines.append("```")
flowchart_text = "\n".join(flowchart_lines)
st.code(flowchart_text, language=None)
# Show step details
st.markdown("**Step Details:**")
for i, step in enumerate(steps, 1):
# Check for errors in this step
step_errors = []
if has_results:
classified_errors = results.get('classified_errors', [])
step_errors = [e for e in classified_errors if e.get('step_number') == i]
if step_errors:
with st.expander(f"Step {i}: {step[:60]}{'...' if len(step) > 60 else ''} ❌", expanded=False):
st.write(f"**Full Step:** {step}")
for error in step_errors:
st.error(f"**Error Found:** {error.get('category', 'Unknown')}")
st.write(f"- Found: `{error.get('found', 'N/A')}`")
st.write(f"- Correct: `{error.get('correct', 'N/A')}`")
explanations = results.get('explanations', {})
if i in explanations:
st.info(f"**Explanation:** {explanations[i]}")
else:
with st.expander(f"Step {i}: {step[:60]}{'...' if len(step) > 60 else ''} ✅", expanded=False):
st.write(f"**Full Step:** {step}")
st.success("No errors detected in this step")
else:
st.info("Enter solution steps to see the problem flowchart")
st.markdown("---")
# Step 1: INPUT
with st.expander("📥 **STEP 1: INPUT** - Problem & Solution Steps", expanded=True):
st.markdown("""
**What happens here?**
- The system receives the math problem and step-by-step solution
- Input is validated and prepared for processing
- Steps are parsed and segmented for analysis
""")
if problem:
st.success(f"✅ Received problem: {problem}")
if steps:
st.success(f"✅ Received {len(steps)} solution steps")
if has_results:
st.code(f"Problem: {results.get('problem', '')[:100]}...")
# Step 2: PARSING
with st.expander("🔍 **STEP 2: PARSING** - Extract Mathematical Expressions", expanded=has_results):
st.markdown("""
**What happens here?**
- Mathematical expressions are extracted using regex patterns
- Operations (+, -, *, /) are identified
- Numbers and variables are recognized
- Each step is prepared for verification
""")
if has_results:
steps = results.get('steps', [])
st.success(f"✅ Parsed {len(steps)} steps")
for i, step in enumerate(steps[:3], 1):
st.write(f" Step {i}: {step[:60]}...")
if len(steps) > 3:
st.write(f" ... and {len(steps) - 3} more steps")
# Step 3: PARALLEL EXECUTION
with st.expander("🔄 **STEP 3: PARALLEL EXECUTION** - 3 Models Running Simultaneously", expanded=has_results):
st.markdown("""
**What happens here?**
- **Model 1 (Symbolic) 🔢**: Uses SymPy to verify all arithmetic calculations
- Weight: 40% (most reliable for math)
- Not affected by sidebar selection
- **Model 2 (LLM Logical) 🧠**: Checks for logical consistency and contradictions
- Weight: 35%
- Uses first selected model from sidebar (e.g., GPT-4)
- Currently: Pattern-based simulation
- **Model 3 (Ensemble) 🤖**: Simulates multiple LLMs voting on solution validity
- Weight: 25%
- Uses ALL selected models from sidebar (GPT-4, Llama 2, Gemini)
- Each model votes, majority wins
- Currently: Pattern-based simulation
All three models run **in parallel** using ThreadPoolExecutor for speed!
""")
if has_results:
model_results = results.get('model_results', {})
col1, col2, col3 = st.columns(3)
with col1:
if 'symbolic' in model_results:
verdict = model_results['symbolic']['verdict']
conf = model_results['symbolic']['confidence'] * 100
errors = len(model_results['symbolic'].get('errors', []))
if verdict == "ERROR":
st.error(f"**🔢 Symbolic**\n\n❌ {verdict}\n\nConfidence: {conf:.1f}%\n\nErrors: {errors}")
else:
st.success(f"**🔢 Symbolic**\n\n✅ {verdict}\n\nConfidence: {conf:.1f}%\n\nErrors: {errors}")
with col2:
if 'llm_logical' in model_results:
verdict = model_results['llm_logical']['verdict']
conf = model_results['llm_logical']['confidence'] * 100
errors = len(model_results['llm_logical'].get('errors', []))
if verdict == "ERROR":
st.error(f"**🧠 LLM Logical**\n\n❌ {verdict}\n\nConfidence: {conf:.1f}%\n\nErrors: {errors}")
else:
st.success(f"**🧠 LLM Logical**\n\n✅ {verdict}\n\nConfidence: {conf:.1f}%\n\nErrors: {errors}")
with col3:
if 'ensemble' in model_results:
verdict = model_results['ensemble']['verdict']
conf = model_results['ensemble']['confidence'] * 100
agreement = model_results['ensemble'].get('agreement', 'N/A')
if verdict == "ERROR":
st.error(f"**🤖 Ensemble**\n\n❌ {verdict}\n\nConfidence: {conf:.1f}%\n\nAgreement: {agreement}")
else:
st.success(f"**🤖 Ensemble**\n\n✅ {verdict}\n\nConfidence: {conf:.1f}%\n\nAgreement: {agreement}")
else:
st.info("⏳ Models will execute in parallel when you click 'Verify Solution'")
# Step 4: CONSENSUS
with st.expander("⚖️ **STEP 4: CONSENSUS** - Weighted Voting Mechanism", expanded=has_results):
st.markdown("""
**What happens here?**
- The system combines results from all 3 models using **weighted voting**:
- **Symbolic Model**: 40% weight (most reliable for arithmetic)
- **LLM Logical Model**: 35% weight (good for reasoning)
- **Ensemble Model**: 25% weight (provides diversity)
- An **error score** is calculated: if > 0.50, verdict = ERROR
- **Confidence** is adjusted based on agreement:
- All 3 agree: confidence boosted by 10%
- 2/3 agree: uses average of agreeing models
- Mixed: confidence penalized by 20%
""")
if has_results:
consensus = results.get('consensus', {})
final_verdict = consensus.get('final_verdict', 'UNKNOWN')
overall_conf = consensus.get('overall_confidence', 0) * 100
error_score = consensus.get('error_score', 0)
agreement = consensus.get('agreement_type', 'UNKNOWN')
st.markdown(f"""
**Consensus Results:**
- **Final Verdict**: {'❌ ERROR' if final_verdict == 'ERROR' else '✅ VALID'}
- **Overall Confidence**: {overall_conf:.1f}%
- **Error Score**: {error_score:.3f} (threshold: 0.50)
- **Agreement Type**: {agreement}
""")
# Show individual model contributions
st.markdown("**Model Contributions:**")
individual_verdicts = consensus.get('individual_verdicts', {})
individual_confidences = consensus.get('individual_confidences', {})
weights = {"symbolic": 0.40, "llm_logical": 0.35, "ensemble": 0.25}
for model_name, verdict in individual_verdicts.items():
weight = weights.get(model_name, 0)
confidence = individual_confidences.get(model_name, 0) * 100
contribution = weight * individual_confidences.get(model_name, 0)
st.write(f" - **{model_name.title()}**: {verdict} ({confidence:.1f}% confidence) → {weight*100:.0f}% weight → contributes {contribution:.3f}")
# Step 5: ERROR CLASSIFICATION
with st.expander("🏷️ **STEP 5: ERROR CLASSIFICATION** - Categorize & Analyze Errors", expanded=has_results and len(results.get('classified_errors', [])) > 0):
st.markdown("""
**What happens here?**
- Each detected error is classified into one of 10+ error types:
- Arithmetic Error (calculation mistakes)
- Logical Error (contradictions)
- Operation Mismatch (says one thing, does another)
- Semantic Error (meaning doesn't match)
- And more...
- **Severity** is assigned: HIGH, MEDIUM, or LOW
- **Fixability** is assessed: can the error be auto-corrected?
""")
if has_results:
classified_errors = results.get('classified_errors', [])
if classified_errors:
st.success(f"✅ Classified {len(classified_errors)} error(s)")
for error in classified_errors[:3]:
st.markdown(f"""
**Error in Step {error.get('step_number', '?')}:**
- **Category**: {error.get('category', 'Unknown')}
- **Severity**: {error.get('severity', 'Unknown')}
- **Fixable**: {'Yes' if error.get('fixable', False) else 'No'}
- **Fixability Score**: {error.get('fixability_score', 0)*100:.0f}%
""")
else:
st.info("✅ No errors found - solution is valid!")
# Step 6: EXPLANATION GENERATION
with st.expander("💬 **STEP 6: EXPLANATION GENERATION** - Create Human-Readable Explanations", expanded=has_results and len(results.get('explanations', {})) > 0):
st.markdown("""
**What happens here?**
- For each error, a natural language explanation is generated
- Explains **why** the error occurred
- Provides educational context
- Suggests how to avoid similar mistakes
- Includes learning tips
""")
if has_results:
explanations = results.get('explanations', {})
if explanations:
st.success(f"✅ Generated {len(explanations)} explanation(s)")
for step_num, explanation in list(explanations.items())[:2]:
with st.container():
st.markdown(f"**Step {step_num} Explanation:**")
st.info(explanation)
else:
st.info("✅ No explanations needed - solution is correct!")
# Step 7: ERROR CORRECTION
with st.expander("🔧 **STEP 7: ERROR CORRECTION** - Automatic Fixes", expanded=has_results and results.get('correction', {}).get('fixed_count', 0) > 0):
st.markdown("""
**What happens here?**
- Fixable errors are automatically corrected
- Arithmetic errors: correct values are calculated and replaced
- Operation mismatches: operations are corrected
- Success rate is tracked for each error type
- Errors requiring manual review are flagged
""")
if has_results:
correction = results.get('correction', {})
fixed_count = correction.get('fixed_count', 0)
if fixed_count > 0:
st.success(f"✅ Fixed {fixed_count} error(s)")
st.write(f"**Success Rate**: {correction.get('success_rate', 0)*100:.1f}%")
correction_log = correction.get('correction_log', [])
if correction_log:
for log_entry in correction_log[:2]:
st.markdown(f"""
**Step {log_entry['step']} ({log_entry['type']}):**
- Original: `{log_entry['original']}`
- Corrected: `{log_entry['corrected']}`
""")
else:
st.info("✅ No corrections needed")
# Step 8: OUTPUT
with st.expander("📤 **STEP 8: OUTPUT** - Final Results", expanded=has_results):
st.markdown("""
**What happens here?**
- Final verdict is displayed (VALID or ERROR)
- Overall confidence score is shown
- All errors with explanations are presented
- Processing time is reported
- Results are ready for review
""")
if has_results:
consensus = results.get('consensus', {})
final_verdict = consensus.get('final_verdict', 'UNKNOWN')
overall_conf = consensus.get('overall_confidence', 0) * 100
processing_time = results.get('processing_time', 0)
total_errors = len(results.get('classified_errors', []))
if final_verdict == "ERROR":
st.error(f"**Final Verdict**: ❌ {final_verdict}")
else:
st.success(f"**Final Verdict**: ✅ {final_verdict}")
st.metric("Overall Confidence", f"{overall_conf:.1f}%")
st.metric("Processing Time", f"{processing_time:.3f}s")
st.metric("Total Errors Found", total_errors)
st.success("✅ Verification complete! Results displayed above.")
else:
st.info("⏳ Results will appear here after verification")
# Visual flowchart diagram
st.markdown("---")
st.markdown("### 📊 Processing Flow Diagram")
st.markdown("""
```
┌─────────────────────────────────────────────────────────┐
│ 📥 INPUT │
│ Problem + Solution Steps │
└────────────────────┬────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────┐
│ 🔍 PARSING │
│ Extract expressions, identify operations │
└────────────────────┬────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────┐
│ 🔄 PARALLEL EXECUTION │
│ ┌──────────┐ ┌──────────┐ ┌──────────┐ │
│ │ Symbolic │ │ LLM │ │ Ensemble │ │
│ │ (SymPy) │ │ Logical │ │ (Voting)│ │
│ │ 40% │ │ 35% │ │ 25% │ │
│ └────┬─────┘ └────┬─────┘ └────┬─────┘ │
└───────┼─────────────┼──────────────┼──────────────────────┘
│ │ │
└─────────────┴──────────────┘
│
▼
┌─────────────────────────────────────────────────────────┐
│ ⚖️ CONSENSUS │
│ Weighted Voting → Final Verdict & Confidence │
└────────────────────┬────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────┐
│ 🏷️ ERROR CLASSIFICATION │
│ Categorize → Severity → Fixability │
└────────────────────┬────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────┐
│ 💬 EXPLANATION GENERATION │
│ Natural language explanations for each error │
└────────────────────┬────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────┐
│ 🔧 ERROR CORRECTION │
│ Auto-fix fixable errors → Track success rate │
└────────────────────┬────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────┐
│ 📤 OUTPUT │
│ Final Verdict + Confidence + All Details │
└─────────────────────────────────────────────────────────┘
```""")
def display_logs():
"""Display processing logs with color coding."""
if not st.session_state.steps_log:
return
st.subheader("📊 Processing Flow")
for log_entry in st.session_state.steps_log:
status = log_entry["status"]
step = log_entry["step"]
model = log_entry["model"]
details = log_entry["details"]
# Color coding based on status
if status.startswith("✓"):
st.success(f"**{status}** [{model}] {step}: {details}")
elif status.startswith("❌"):
st.error(f"**{status}** [{model}] {step}: {details}")
elif status.startswith("⚠️"):
st.warning(f"**{status}** [{model}] {step}: {details}")
else:
st.info(f"**{status}** [{model}] {step}: {details}")
def display_results():
"""Display verification results."""
if not st.session_state.results:
return
results = st.session_state.results
st.header("🎯 Results")
# Final verdict, confidence, processing time
col1, col2, col3 = st.columns(3)
consensus = results.get("consensus", {})
final_verdict = consensus.get("final_verdict", "UNKNOWN")
overall_confidence = consensus.get("overall_confidence", 0.0)
processing_time = results.get("processing_time", 0.0)
with col1:
if final_verdict == "ERROR":
st.error(f"**Final Verdict:** ❌ {final_verdict}")
else:
st.success(f"**Final Verdict:** ✅ {final_verdict}")
with col2:
st.metric("**Confidence**", f"{overall_confidence * 100:.1f}%")
with col3:
st.metric("**Processing Time**", f"{processing_time:.2f}s")
# Model verdicts
st.subheader("🤖 Model Verdicts")
model_results = results.get("model_results", {})
cols = st.columns(3)
model_names = ["symbolic", "llm_logical", "ensemble"]
model_display_names = {
"symbolic": "🔢 Symbolic",
"llm_logical": "🧠 LLM Logical",
"ensemble": "🤖 Ensemble"
}
for idx, model_name in enumerate(model_names):
with cols[idx]:
if model_name in model_results:
model_result = model_results[model_name]
verdict = model_result.get("verdict", "UNKNOWN")
confidence = model_result.get("confidence", 0.0)
errors_count = len(model_result.get("errors", []))
if verdict == "ERROR":
st.error(f"**{model_display_names[model_name]}**\n\nVerdict: ❌ {verdict}\n\nConfidence: {confidence * 100:.1f}%\n\nErrors: {errors_count}")
else:
st.success(f"**{model_display_names[model_name]}**\n\nVerdict: ✅ {verdict}\n\nConfidence: {confidence * 100:.1f}%\n\nErrors: {errors_count}")
# Consensus mechanism breakdown
st.subheader("⚖️ Consensus Mechanism")
agreement_type = consensus.get("agreement_type", "UNKNOWN")
error_score = consensus.get("error_score", 0.0)
individual_verdicts = consensus.get("individual_verdicts", {})
individual_confidences = consensus.get("individual_confidences", {})
st.info(f"**Agreement:** {agreement_type}")
st.info(f"**Error Score:** {error_score:.3f} (threshold: 0.50)")
st.write("**Individual Model Results:**")
for model_name, verdict in individual_verdicts.items():
confidence = individual_confidences.get(model_name, 0.0)
st.write(f"- {model_display_names.get(model_name, model_name)}: {verdict} (confidence: {confidence * 100:.1f}%)")
# Error details
classified_errors = results.get("classified_errors", [])
if classified_errors:
st.subheader("🔴 Error Details")
for error in classified_errors:
with st.expander(f"Error in Step {error.get('step_number', 0)}: {error.get('category', 'Unknown')}"):
st.write(f"**Type:** {error.get('type', 'unknown')}")
st.write(f"**Found:** {error.get('found', 'N/A')}")
st.write(f"**Correct:** {error.get('correct', 'N/A')}")
st.write(f"**Severity:** {error.get('severity', 'UNKNOWN')}")
st.write(f"**Fixable:** {'Yes' if error.get('fixable', False) else 'No'}")
# Show explanation
explanations = results.get("explanations", {})
step_num = error.get("step_number", 0)
if step_num in explanations:
st.write(f"**Explanation:** {explanations[step_num]}")
# Correction results
correction = results.get("correction", {})
if correction and correction.get("fixed_count", 0) > 0:
st.subheader("🔧 Corrections Applied")
st.success(f"**Fixed:** {correction.get('fixed_count', 0)} / {correction.get('total_fixable', 0)} errors")
st.write(f"**Success Rate:** {correction.get('success_rate', 0.0) * 100:.1f}%")
correction_log = correction.get("correction_log", [])
if correction_log:
with st.expander("View Correction Log"):
for log_entry in correction_log:
st.write(f"**Step {log_entry['step']}:** {log_entry['type']}")
st.write(f"Original: {log_entry['original']}")
st.write(f"Corrected: {log_entry['corrected']}")
# Main UI
st.set_page_config(page_title="Math Verification System", page_icon="🔢", layout="wide")
st.title("🔢 Mathematical Reasoning Verification System")
st.markdown("3-Model Parallel Verification with Weighted Consensus")
# Sidebar configuration
with st.sidebar:
st.header("⚙️ Configuration")
gpt4_enabled = st.checkbox("GPT-4", value=True)
llama_enabled = st.checkbox("Llama 2", value=True)
gemini_enabled = st.checkbox("Gemini", value=True)
selected_models = []
if gpt4_enabled:
selected_models.append("GPT-4")
if llama_enabled:
selected_models.append("Llama 2")
if gemini_enabled:
selected_models.append("Gemini")
if not selected_models:
selected_models = ["GPT-4", "Llama 2", "Gemini"]
st.info(f"Selected models: {', '.join(selected_models)}")
st.markdown("---")
st.markdown("### 📖 How Models Are Used")
with st.expander("ℹ️ Click to see how sidebar models work"):
st.markdown("""
**Model 1 (Symbolic) 🔢:**
- Uses SymPy library (not affected by sidebar)
- Verifies arithmetic calculations
- Weight: 40%
**Model 2 (LLM Logical) 🧠:**
- Uses first selected model from sidebar
- Checks logical consistency
- Weight: 35%
**Model 3 (Ensemble) 🤖:**
- Uses ALL selected models from sidebar
- Each model votes on solution validity
- Majority voting determines verdict
- Weight: 25%
**Note:** Currently using pattern-based simulation.
For production, integrate real LLM APIs (OpenAI, Anthropic, Google).
""")
st.markdown("### 🔄 Model Flow Diagram")
st.code("""
Sidebar Selection
┌─────────────────┐
│ GPT-4 ✓ │
│ Llama 2 ✓ │──┐
│ Gemini ✓ │ │
└─────────────────┘ │
│
┌─────────────────────┴─────────────────────┐
│ │
▼ ▼
Model 2 (LLM Logical) Model 3 (Ensemble)
Uses: GPT-4 (first selected) Uses: All selected
Weight: 35% Weight: 25%
└───────────────────────────────────────────┘
│
▼
Consensus Mechanism
(Weighted Voting)
""", language=None)
# Main layout
col_left, col_right = st.columns([1, 1])
with col_left:
st.header("📝 Input")
problem = st.text_area(
"Problem:",
height=80,
placeholder="Enter the math problem here...",
value="Janet has 3 apples. She buys 2 more. She gives 1 away. How many?"
)
steps_input = st.text_area(
"Solution Steps (one per line):",
height=120,
placeholder="Enter solution steps, one per line...",
value="Janet starts with 3 apples\nShe buys 2 more: 3 + 2 = 5 apples\nShe gives 1 away: 5 - 1 = 6 apples"
)
col_btn1, col_btn2 = st.columns(2)
with col_btn1:
verify_button = st.button("🚀 Verify Solution", type="primary", use_container_width=True)
with col_btn2:
clear_button = st.button("🔄 Clear", use_container_width=True)
with col_right:
st.header("🎯 Live Flowchart")
display_flowchart(problem=problem, steps_input=steps_input)
# Handle button clicks
if verify_button:
# Clear previous logs
st.session_state.steps_log = []
st.session_state.results = None
# Parse steps
steps = [s.strip() for s in steps_input.split('\n') if s.strip()]
if not problem or not steps:
st.warning("Please enter both problem and solution steps.")
else:
# Add initial log
add_log("Verification Started", "System", "⏳", "Initializing models...")
# Run verification
with st.spinner("Running verification..."):
try:
# Add logs for each model starting
add_log("Model 1 Started", "Symbolic", "🔢", "Checking arithmetic...")
add_log("Model 2 Started", "LLM Logical", "🧠", "Checking logical consistency...")
add_log("Model 3 Started", "Ensemble", "🤖", "Running ensemble voting...")
# Use first selected model for LLM Logical, or default to GPT-4
llm_model_name = selected_models[0] if selected_models else "GPT-4"
results = run_verification_parallel(
problem=problem,
steps=steps,
model_name=llm_model_name,
model_list=selected_models
)
# Add completion logs
for model_name, model_result in results["model_results"].items():
verdict = model_result.get("verdict", "UNKNOWN")
errors_count = len(model_result.get("errors", []))
status = "✓ ERROR" if verdict == "ERROR" else "✓ VALID"
add_log(
f"Model {model_name} Completed",
model_name.title(),
status,
f"Found {errors_count} error(s)"
)
# Add consensus log
consensus_verdict = results["consensus"].get("final_verdict", "UNKNOWN")
add_log(
"Consensus Computed",
"Consensus",
"⚖️",
f"Final verdict: {consensus_verdict}"
)
st.session_state.results = results
except Exception as e:
st.error(f"Error during verification: {str(e)}")
add_log("Error", "System", "❌", str(e))
if clear_button:
st.session_state.steps_log = []
st.session_state.results = None
st.rerun()
# Display logs and results
display_logs()
display_results()
|