LetTokenBurnA Corpus and Benchmark for Nonlinear Narrative Generation
Evaluates LLMs on generating nonlinear stories from 83 annotated works using event lists and story graphs across seven narrative shapes.
$ npm i -g lettokenburn-cli$ ltb bench info nonlinear-narrative-generation-benchmark
External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| Gemini 3.6 Flash | 0.8 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
| Gemini 3.1 Pro | 0.7 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
| Grok 4.5 | 0.7 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
| GPT-5.6 Sol | 0.7 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
| GPT-5.6-terra | 0.6 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
| Claude Opus 5 | 0.6 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
| GLM-5.2 | 0.6 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
| Nemotron 3 Ultra | 0.5 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
| GPT-5.6 Luna | 0.5 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
| Claude Haiku 4.5 | 0.5 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
| DeepSeek V4 Pro | 0.5 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
| Claude Sonnet 5 | 0.4 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
| Llama 4 Maverick | 0.3 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
| GPT-oss 120B | 0.3 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.
The benchmark comprises 83 published works annotated with chronological/telling orders and story graphs. It evaluates 19 models on five tasks involving narrative structure control.
Scoring uses automated judges and human preference ratings. Models struggle to produce preferred nonlinear narratives, often defaulting to chronological order despite instructions.