Commit History

Update Colab notebook: 1.5B model, scaled rewards, tuned hyperparameters
ee8c2d4
Running

nihalaninihal commited on

Fix critical RL reward function exploits and training hyperparameters
803c93e

nihalaninihal Claude Opus 4.6 commited on

Align with Advanced Llama 3.2 GRPO LoRA reference notebook pattern
c7d253a

nihalaninihal Claude Opus 4.6 commited on

Fix VALID_TARGETS_FOR_ATTACK and attacker heuristic/prompt inconsistencies
3ffb78a

nihalaninihal Claude Opus 4.6 commited on

Fix format_comparison_metrics_html to accept run_comparison() dict directly
d52b449

nihalaninihal Claude Opus 4.6 commited on

Add run_demo_episode wrapper to demo.py for dict-based episode results
fcf34b9

nihalaninihal Claude Opus 4.6 commited on

Fix Gradio 6 deprecation warning: move theme/css out of Blocks constructor
af292c9

nihalaninihal commited on

Align train.py and Colab notebook with official Unsloth+OpenEnv GRPO patterns
e09a415

nihalaninihal Claude Opus 4.6 commited on

Update metrics format with drift/oversight tracking, add colab training notebook
5e0f2b1

nihalaninihal commited on

Fix requirements: add pandas>=2.0, set gradio>=6.0.0 consistently
173a3e9

nihalaninihal commited on

Fix theme in Blocks constructor for HF Spaces compatibility
ca88c10

nihalaninihal commited on

Add billing schema drift support with field renaming
500061c

nihalaninihal commited on

Add drift-specific metrics: drift events, detection, adaptation rate
eb9e808

nihalaninihal commited on

Add oversight accuracy and explanation quality metrics to dashboard
33b6c02

nihalaninihal commited on

Improve HeuristicOversight explanations with specific data references
62aabbf

nihalaninihal commited on

Add SLA policy drift support for ticketing system
5014574

nihalaninihal commited on

Add structured explanation quality scoring for oversight agent
5efcc1b

nihalaninihal commited on

Fix schema drift renames to target actual Customer model fields
197e7c5

nihalaninihal commited on

Fix window_ticks policy enforcement in billing refund validation
aea9d7d

nihalaninihal commited on

Add master improvement plan with prioritized fixes for hackathon submission
7f33a54

nihalaninihal commited on

Add multi-agent GRPO training for all 3 agents (worker, attacker, oversight)
389e3bf

nihalaninihal Claude Opus 4.6 commited on

Add comprehensive gap analysis and 4-hour action plan for hackathon submission
ea3624f

nihalaninihal Claude Opus 4.6 commited on

Add randomized attacker, security metrics engine, and updated Gradio dashboard
69a7e43

nihalaninihal Claude Opus 4.6 commited on

Add episode metrics computation and HTML formatting for SentinelOps Arena
23f3257

nihalaninihal Claude Opus 4.6 commited on

Replace HeuristicAttacker with RandomizedAttacker for probabilistic attacks
1f6f2a5

nihalaninihal Claude Opus 4.6 commited on

Improve Gradio UI layout with sidebar controls, sub-tabs, and styled score widgets
e85e584

nihalaninihal Claude Opus 4.6 commited on

Revamp Gradio app with Gradio 6, custom cybersecurity theme, and rich visualizations
f20603d

nihalaninihal Claude Opus 4.6 commited on

Remove hackathon_env template, rewrite train.py for SentinelOpsArena
0e5a0a6

nihalaninihal Claude Opus 4.6 commited on

Implement Phase 3 (HTTP server) and Phase 4 (demo + Gradio app)
fa00f5a

nihalaninihal Claude Opus 4.6 commited on

Implement Phase 2: environment core with MCPEnvironment base
6c20e91

nihalaninihal Claude Opus 4.6 commited on

Implement Phase 1: models, enterprise systems, attacks, rewards
a4e6593

nihalaninihal Claude Opus 4.6 commited on

Refine build plan with devil's advocate corrections
dc8bc66

nihalaninihal Claude Opus 4.6 commited on

Add phased build plan and setup guide for SentinelOps Arena
707377e

nihalaninihal Claude Opus 4.6 commited on

Update SentinelOps Arena with detailed 14-hour implementation plan
5f590b1

nihalaninihal Claude Opus 4.6 commited on

Add SentinelOps Arena project specification
af942b1

nihalaninihal Claude Opus 4.6 commited on

Initial project setup for OpenEnv Hackathon
ccb5f4e

nihalaninihal Claude Opus 4.6 commited on