IG-Bench
Models trace scientific idea lineages across 1,029 questions and propose research ideas grounded in those lineages. IG-Exam requires every answer field to match.
IG-Arena uses a separate judge-panel score for heredity, variation, and selection value. The paper calls the combined evaluation IG-Bench.
- Released
- July 2026
- Kind
- Research science
- Level
- Research frontier
- Scored
- Checked automatically
- Answers
- Short answer
- Tasks
- All public
- Measures
- Capability
- Subject
- Computer and information sciences, Physical sciences, Chemical sciences, Biological sciences
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Biology and medicine, Chemistry and materials
- Strongest model tested
- GPT-5.5 #21 on the capability index
- Last model added
- April 2026
- Scores refreshed
- November 20th, 2026
In the authors’ words
Scientific ideas rarely start from a blank page. They inherit mechanisms, repair known limitations, and recombine pieces of earlier work, much like biological genomes. Current benchmarks still say little about whether AI systems can follow this inheritance structure. We present IdeaGene-Bench (IG-Bench), a benchmark for scientific lineage reasoning and lineage-grounded idea generation. IG-Bench is organized around the IdeaGene framework: each paper or proposal is represented as a set of minimal, typed, evidence-grounded Idea Genome objects, and a GenomeDiff aligns these objects to record inheritance, mutation, loss, external import, and novel insertion under six operational evolutionary dynamics. The benchmark contains 1,961 golden lineage traces, 1,085 curated Idea Genome objects, and 920 pairwise GenomeDiff records across 10 scientific domains. It supports two evaluations. IG-Exam (42 task types, 1,029 instances) tests closed-form lineage reasoning across Idea Genome abstraction, inheritance tracing, evolutionary reasoning, and lineage verification. IG-Arena evaluates generation with a lineage-conditioned Population-Evolution Score(PES), asking whether a proposal can be inserted as a coherent descendant of a given lineage population: it should inherit the right Idea Genome objects, vary meaningfully from nearby work, and offer selection value for future research. Experiments on 14 LLM-based scientists expose a compositional bottleneck. The strongest system reaches only 27.3% exact accuracy on lineage reasoning, and structured lineage context reshuffles system rankings rather than helping every participant uniformly.
GPT-5.5 leads with 27.3% among models released by April 2026
Best score per model from the benchmark's own leaderboard, read on November 20th, 2026. The authors report each track separately. The headline shows total IG-Exam exact accuracy. Where a model ran with several settings or harnesses, we keep its best.
The newest model with a score here came out in April 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1 GPT-5.5 27.3%
- 2 Claude Opus 4.7 22%
- 3
Qwen 3.6 Max (Preview) 20.6%
- 4 Kimi K2 Thinking 20.4%
- 5
Gemini 3.1 Pro 20.1%
- 6
DeepSeek-V4-Pro 17.9%
- 7 GLM-5.1 17.7%
- 8
MiniMax-M2.7 11.9%
Source: Official IG-Bench leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-5.5 | 27.3% | Claude Code | Apr 2026 | Official leaderboard |
| 2 | Claude Opus 4.7 | 22% | Claude Code | Apr 2026 | Official leaderboard |
| 3 | | 20.6% | Apr 2026 | Official leaderboard | |
| 4 | Kimi K2 Thinking | 20.4% | Nov 2025 | Official leaderboard | |
| 5 | | 20.1% | Feb 2026 | Official leaderboard | |
| 6 | | 17.9% | Apr 2026 | Official leaderboard | |
| 7 | GLM-5.1 | 17.7% | Apr 2026 | Official leaderboard | |
| 8 | | 11.9% | Mar 2026 | Official leaderboard |
Rankings by part
We read 6 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as T1 Abstraction, T2 Tracing and T3 Reasoning. GPT-5.5 leads 4 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
T1 Abstraction
8 models we track have a score here.
- 1 Claude Opus 4.7 ↑1 34.4%
- 2
Gemini 3.1 Pro ↑3 32.4%
- 3 GPT-5.5 ↓2 31.8%
- 4 GLM-5.1 ↑3 28.5%
- 5 Kimi K2 Thinking ↓1 27.9%
- 6
Qwen 3.6 Max (Preview) ↓3 26.9%
- 7
DeepSeek-V4-Pro ↓1 23.9%
- 8
MiniMax-M2.7 22.9%
T2 Tracing
8 models we track have a score here.
- 1 GPT-5.5 37.9%
- 2 Claude Opus 4.7 29.7%
- 3
Gemini 3.1 Pro ↑2 24.6%
- 4
Qwen 3.6 Max (Preview) ↓1 22.5%
- 5 Kimi K2 Thinking ↓1 22.1%
- 6
DeepSeek-V4-Pro 20.6%
- 7 GLM-5.1 19.7%
- 8
MiniMax-M2.7 9.10%
T3 Reasoning
8 models we track have a score here.
- 1 GPT-5.5 25.3%
- 2 Claude Opus 4.7 21%
- 3 Kimi K2 Thinking ↑1 19.6%
- 4
Qwen 3.6 Max (Preview) ↓1 18.8%
- 5
DeepSeek-V4-Pro ↑1 18.6%
- 6
Gemini 3.1 Pro ↓1 17.8%
- 7 GLM-5.1 17.4%
- 8
MiniMax-M2.7 10.9%
T4 Verification
8 models we track have a score here.
- 1
Qwen 3.6 Max (Preview) ↑2 17.4%
- 2 GPT-5.5 ↓1 16%
- 3 Kimi K2 Thinking ↑1 14.8%
- 4 Claude Opus 4.7 ↓2 14.5%
- 5
MiniMax-M2.7 ↑3 11.6%
- 6
Gemini 3.1 Pro ↓1 10.7%
- 7 GLM-5.1 8.50%
- 8
DeepSeek-V4-Pro ↓2 8.50%
IG-Arena PES
8 models we track have a score here.
- 1 GPT-5.5 86.7
- 2 Claude Opus 4.7 83.3
- 3
DeepSeek-V4-Pro ↑3 82.7
- 4
Gemini 3.1 Pro ↑1 82.1
- 5 Kimi K2 Thinking ↓1 81.1
- 6 GLM-5.1 ↑1 80.7
- 7
Qwen 3.6 Max (Preview) ↓4 79.8
- 8
MiniMax-M2.7 70.5
IG-Arena Elo
8 models we track have a score here.
- 1 GPT-5.5 1,476
- 2 Claude Opus 4.7 1,160
- 3
DeepSeek-V4-Pro ↑3 919
- 4
Qwen 3.6 Max (Preview) ↓1 883
- 5 Kimi K2 Thinking ↓1 794
- 6
Gemini 3.1 Pro ↓1 706
- 7 GLM-5.1 668
- 8
MiniMax-M2.7 541
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official IG-Bench leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official IG-Bench leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Science and reasoning
- GPQA Diamond In use 198 graduate-level 'Google-proof' multiple-choice questions in biology, chemistry and physics.
Gemini 4 Carbon 97%
- Terminal-Bench-Science In use 70 scientific research-workflow tasks across five science areas, run in Harbor containers. Hidden tests give the resolution rate. GPT-6 Astra 68.1%
- BioMysteryBench In use Bioinformatics problems on anonymized real datasets that require real analysis with tools. Rubric-graded, with human-solvable and human-difficult splits. Claude Fable 5.2 83.7%
- FrontierScience In use Models solve physics, chemistry, and biology problems in an Olympiad track graded by short answer and a Research track graded by 10-point rubrics.
Intern-S2-397B 87%
- HiPhO In use Models solve theory problems from 2024 and 2025 physics olympiad exams, some with diagrams. Official marking schemes grade answers and steps, mapped to medal levels.
Gemini 3 Pro 78.3%
- LABBench2 In use 11 biology research sub-tasks (literature, figures, tables, patents, trials and more) with tool access. Macro-averaged accuracy. GPT-5.2 60.9%
Data audit
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 7 months ago, in April 2026, for DeepSeek-V4-Pro. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the official leaderboard (8 entries).
- Pass Models scored 8 models we track.
- Warning Newest score The newest came 7 months ago, in April 2026, for DeepSeek-V4-Pro.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.5 (#21).
- Pass Headroom The best is 27.3%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on visionxlab.github.io: 14 rows for models we track at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.