Datasets:

violetxi
/

ch-trajectory-pool-qwen3.5-4b

rollout_id stringlengths 8 15	mode stringclasses 3 values	status stringclasses 4 values	n_turns int64 3 20	n_tool_calls int64 0 20	n_malformed int64 0 7	n_tool_errors int64 0 7	prompt_tokens int64 1.55k 664k	completion_tokens int64 401 38.9k	seed_ref stringlengths 4 117	usable bool 2 classes	task stringlengths 327 2.86k	messages listlengths 7 42
b1c_1001	B1C	done	18	16	1	2	446,054	15,165	1001	true	"Survey client 1001 across the firm's DMS. All of this client's matters live under matters/1001-*. I(...TRUNCATED)	[{"role":"system","content":"You are an agent exploring the document management system (DMS) of the (...TRUNCATED)
b1c_1002	B1C	done	8	6	1	0	108,684	9,371	1002	true	"Survey client 1002 across the firm's DMS. All of this client's matters live under matters/1002-*. I(...TRUNCATED)	[{"role":"system","content":"You are an agent exploring the document management system (DMS) of the (...TRUNCATED)
b1c_1003	B1C	done	15	14	0	1	351,491	6,710	1003	true	"Survey client 1003 across the firm's DMS. All of this client's matters live under matters/1003-*. I(...TRUNCATED)	[{"role":"system","content":"You are an agent exploring the document management system (DMS) of the (...TRUNCATED)
b1c_1004	B1C	done	9	8	0	0	107,483	7,136	1004	true	"Survey client 1004 across the firm's DMS. All of this client's matters live under matters/1004-*. I(...TRUNCATED)	[{"role":"system","content":"You are an agent exploring the document management system (DMS) of the (...TRUNCATED)
b1c_1005	B1C	done	14	12	1	0	337,421	11,345	1005	true	"Survey client 1005 across the firm's DMS. All of this client's matters live under matters/1005-*. I(...TRUNCATED)	[{"role":"system","content":"You are an agent exploring the document management system (DMS) of the (...TRUNCATED)
b1c_1006	B1C	done	13	12	0	0	251,908	8,163	1006	true	"Survey client 1006 across the firm's DMS. All of this client's matters live under matters/1006-*. I(...TRUNCATED)	[{"role":"system","content":"You are an agent exploring the document management system (DMS) of the (...TRUNCATED)
b1c_1007	B1C	done	14	12	1	0	300,385	14,934	1007	true	"Survey client 1007 across the firm's DMS. All of this client's matters live under matters/1007-*. I(...TRUNCATED)	[{"role":"system","content":"You are an agent exploring the document management system (DMS) of the (...TRUNCATED)
b1c_1008	B1C	done	13	12	0	0	285,475	9,347	1008	true	"Survey client 1008 across the firm's DMS. All of this client's matters live under matters/1008-*. I(...TRUNCATED)	[{"role":"system","content":"You are an agent exploring the document management system (DMS) of the (...TRUNCATED)
b1c_1009	B1C	done	12	11	0	0	265,753	9,665	1009	true	"Survey client 1009 across the firm's DMS. All of this client's matters live under matters/1009-*. I(...TRUNCATED)	[{"role":"system","content":"You are an agent exploring the document management system (DMS) of the (...TRUNCATED)
b1c_1010	B1C	done	9	8	0	0	147,117	8,341	1010	true	"Survey client 1010 across the firm's DMS. All of this client's matters live under matters/1010-*. I(...TRUNCATED)	[{"role":"system","content":"You are an agent exploring the document management system (DMS) of the (...TRUNCATED)

End of preview. Expand in Data Studio

C&H Trajectory Pool — Qwen3.5-4B (on-policy)

1,376 agentic exploration trajectories over the full Calderwood & Harkness (C&H) synthetic law-firm corpus (266 matters, ~145M tokens; the open-sourced world from harvey-labs tasks/firm-knowledge/, MIT), generated by Qwen/Qwen3.5-4B — the same model intended as the training student, so this pool is exactly on-policy for it. Part of a world-internalization research project: which likelihood targets, derived from agent experience, internalize a world into model weights.

Companion datasets under this namespace: ch-pilot-rollouts-qwen3.5-35b-a3b and ch-pilot-rollouts-qwen3.5-9b (20-rollout actor-selection pilot).

Task mix (1,376 = one identical, seed-matched spec set)

B1 (266) — one survey per matter: client, parties, timeline, events, status, key documents; path citations required for every claim.
B1C (46) — one survey per client: cross-matter dossier.
B2 (1,064) — self-QA, 4 per matter: given a 2,000-char excerpt from a sampled seed document, pose 3 nontrivial questions (single-doc / cross-doc / cross-matter) and answer them with tools; answers must cite at least one non-seed file, and the cross-matter answer files from ≥2 matters.

Protocol

Text-based tools (glob / grep / read) over a markdown mirror of the corpus, one call per turn as <tool>{json}</tool>, observations returned in <tool_result> (16,000 chars max), answers in <final>...</final>, temperature 0.7, ≤20 turns, 65,536-token context. Generation was mechanically gated (this matters for data quality):

a </tool> stop-sequence ends each turn at the tool call (no runaway generation);
a <final> is accepted only if closed, ≥50 chars, and after ≥1 successful tool call — otherwise the driver returns a corrective error (3 strikes → protocol_failure);
near the context budget (~50k tokens) or turn cap, tool access ends and the model is told to write its final from what it already read (graceful wrap-up instead of overflow loss).

Composition

status	n
done (all usable: ≥1 tool call, closed substantive final)	1,353 (98.3%)
protocol_failure	14
max_turns	7
api_error	2

Each row: full message history (messages: system / user / assistant / tool), run metadata (mode, status, turn/call/error counts, token usage, seed_ref), and a precomputed usable flag. Assistant turns include the model's reasoning text.

Provenance & license

All document content is synthetic (no real firm, clients, or people), derived from Harvey AI's MIT-licensed harvey-labs release (commit 7be41d5); trajectories released under MIT. The 250 firm-knowledge eval tasks are not used or contained here — they remain a held-out evaluation set.

Downloads last month: 32