---
title: README
emoji: 🔍
colorFrom: gray
colorTo: gray
sdk: static
pinned: false
---

# TraceLite AI

An academic-driven open research team working on the building blocks of AI agent evaluation and post-training: **forensics, attribution, environments, rewards, and data generation**. Each component can be used on its own for model and harness selection, training, and pre-release acceptance testing, and together they form a framework for continual self-improvement.

[Website](https://tracelite.ai) · [GitHub](https://github.com/TraceLite-AI) · [Research timeline](https://tracelite.ai/timeline.html) · contact@tracelite.ai

## Trial by Trace · TraceTrial-Bench

When an AI agent fails, was it the model or the harness around it? TraceTrial-Bench scores execution traces in two layers, task outcome and harness reliability, across 95 mechanism-oriented tasks, 4 harnesses (Claude Code, Codex, OpenClaw, DeepSeek Harness) and 9 models, 3,420 runs in total.

- 🔍 [Interactive report and leaderboard (Space)](https://huggingface.co/spaces/TraceLite-AI/Trial-by-Trace)
- 💻 [Code and tasks](https://github.com/TraceLite-AI/TraceTrial-Bench)
- 📄 [Technical report](https://tracelite.ai/trial-by-trace.html)

## Papers (2026)

| Paper | Venue |
|---|---|
| [MM-VeriRubric: Verifiable Evidence-Grounded Rubrics for Multimodal Reward Models](https://tracelite.ai/papers.html) | EMNLP 2026 |
| [Disentangled Privileged States for Simulator-Assisted Multi-Turn Dialogue Policy Training](https://tracelite.ai/papers.html) | EMNLP 2026 Findings |
| [Absorbing Gradient Conflicts: Modeling Semantic Variance via Kent Distributions for Cross-Modal Hashing](https://arxiv.org/abs/2608.24010) | IJCAI 2026 |
| [SimRPD: Optimizing Recruitment Proactive Dialogue Agents through Simulator-Based Data Evaluation and Selection](https://aclanthology.org/2026.acl-industry.95/) | ACL 2026 · Oral |
| [Look Less, Reason More: Rollout-Guided Adaptive Pixel-Space Reasoning](https://aclanthology.org/2026.acl-long.225/) | ACL 2026 |
| [Select Less, Reason More: Prioritizing Evidence Purity for Video Reasoning](https://openaccess.thecvf.com/content/CVPR2026/html/Li_Select_Less_Reason_More_Prioritizing_Evidence_Purity_for_Video_Reasoning_CVPR_2026_paper.html) | CVPR 2026 |
| [CausalStep: A Benchmark for Explicit Stepwise Causal Reasoning in Videos](https://arxiv.org/abs/2507.16878) | AAAI 2026 · Oral |

Full list: [tracelite.ai/papers.html](https://tracelite.ai/papers.html)