TigerGraph Agentic GraphRAG Hackathon

Don't pay an agent to answer what a graph already knows.

OCCAM runs RAG, GraphRAG and Agentic GraphRAG on the same 100 questions, then adds a router that sends each question to the cheapest tier that can truly answer it. Same accuracy as the agent, a fraction of the tokens.

▶ Watch the four pipelines race Source
The pain point

Everyone is adding agents. Nobody is asking what they cost.

The hackathon guidebook asks whether the extra reasoning and retrieval steps are worth the extra complexity and token cost. Most submissions will report accuracy and stop there.

63%

Plain RAG fails structurally

"How many cycling events had more than 30 competitors?" needs up to 43 documents. A top-10 retrieval cannot hold that evidence, so aggregation accuracy is 10%.

97%

GraphRAG stops at ambiguity

When the graph returns a shortlist instead of an answer, a single pass gives up. It lost 3 of 100 questions that way.

890

Agents pay on every question

Agentic GraphRAG reaches 100%, but spends ~890 tokens per question, including the 97 that did not need it.

Live demo

Same question. Four pipelines. Watch the tokens.

Pick a real benchmark question and run it. Each lane replays the recorded trace of that pipeline: every step, every token, every verdict, straight from artifacts/results_public.jsonl.

Replay of recorded runs, not a fresh LLM call: step durations are the measured latencies from the benchmark.

The tradeoff

Up and to the left is better.

Accuracy against mean tokens per question, log scale. The agent reaches the top. Routing reaches the same height for ~14× fewer tokens.

Accuracy by question type

RAG collapses on aggregation (10%) and superlatives (60%).

How OCCAM works

Escalate only when a tier can name what it was missing.

Escalation is never a guess about difficulty. Every tier reports its gap, and that gap is recorded in the trace, so the controller cannot silently skip evidence.

Built to be believed

What we measured, and what we didn't claim.

Accuracy and completeness

We report answer accuracy and evidence recall separately. RAG's 63% against 76% recall is the whole argument for doing both.

Runs on TigerGraph

Every agent tool is an installed GSQL query on Savanna. 78 of 100 questions were verified identical to the reference store, 0 mismatches.

Honest limits

Tier 0 is a rule cache. On six paraphrased questions no rule matches, tier 1 answered all six correctly. Single model, single run, no variance bars.