Datasets:
agent stringclasses 1
value | benchmark stringclasses 1
value | context dict | context_length_error bool 1
class | exception dict | finished_at stringlengths 27 27 | harness_fallback_turn_count int64 0 21 | malformed_assistant_turn_count int64 0 0 | max_completion_tokens_per_turn int64 0 4.47k | max_prompt_tokens_per_turn int64 0 32.7k | message_schema_version stringclasses 1
value | messages listlengths 1 209 | messages_without_thinking listlengths 1 209 | model stringclasses 1
value | model_source stringclasses 1
value | n_agent_turns int64 0 104 | n_cache_tokens int64 0 0 | n_input_tokens int64 0 584k | n_output_tokens int64 0 60.9k | n_test_cases_passed int64 0 9 | n_test_cases_total int64 0 11 | official_error_zero bool 2
classes | output_length_cap_events int64 0 0 | raw_reward float64 0 1 ⌀ | reward float64 0 1 | sampling dict | source_run stringclasses 1
value | started_at stringlengths 27 27 | steps listlengths 1 109 | steps_without_thinking listlengths 1 109 | summarization_count int64 0 43 | task stringclasses 89
values | test_cases listlengths 0 11 | thinking_markup_complete bool 1
class | thinking_markup_policy stringclasses 1
value | thinking_open_tag_inferred bool 2
classes | thinking_turn_count int64 0 104 | trajectory_available bool 1
class | trajectory_integrity dict | trial_id stringlengths 16 41 | untagged_assistant_turn_count int64 0 0 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
terminus-2 | terminal-bench/terminal-bench-2-1 | {"interleaved_thinking":false,"model_context_length":65536,"proactive_summarization_threshold":40960(...TRUNCATED) | false | null | 2026-07-30T09:44:32.537032Z | 21 | 0 | 1,850 | 32,658 | 1.1 | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [REDACTED] | Qwen/Qwen3.5-4B | 47 | 0 | 189,732 | 8,344 | 1 | 2 | false | 0 | 0 | 0 | {"enable_thinking":true,"extra_body":{"chat_template_kwargs":{"enable_thinking":true},"min_p":0.0,"r(...TRUNCATED) | [REDACTED]_base-thinking-canonical-repaired-20260730 | 2026-07-30T09:30:41.342516Z | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | 43 | extract-elf | [{"name":"test_outputs.py::test_extract_js_exists","status":"passed"},{"name":"test_outputs.py::test(...TRUNCATED) | true | preserve-zero-tag-generations-without-inference | true | 47 | true | {"base_present":true,"contiguous":false,"continuation_indices":[1,43],"missing_continuation_indices"(...TRUNCATED) | extract-elf__aP9M3oB | 0 |
terminus-2 | terminal-bench/terminal-bench-2-1 | {"interleaved_thinking":false,"model_context_length":65536,"proactive_summarization_threshold":40960(...TRUNCATED) | false | null | 2026-07-30T09:09:41.741316Z | 20 | 0 | 3,092 | 31,584 | 1.1 | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [REDACTED] | Qwen/Qwen3.5-4B | 49 | 0 | 178,615 | 22,436 | 0 | 1 | false | 0 | 0 | 0 | {"enable_thinking":true,"extra_body":{"chat_template_kwargs":{"enable_thinking":true},"min_p":0.0,"r(...TRUNCATED) | [REDACTED]_base-thinking-canonical-repaired-20260730 | 2026-07-30T08:49:59.804034Z | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | 41 | feal-linear-cryptanalysis | [
{
"name": "test_outputs.py::test_attack",
"status": "failed"
}
] | true | preserve-zero-tag-generations-without-inference | true | 49 | true | {"base_present":true,"contiguous":false,"continuation_indices":[1,41],"missing_continuation_indices"(...TRUNCATED) | feal-linear-cryptanalysis__TcnfzNy | 0 |
terminus-2 | terminal-bench/terminal-bench-2-1 | {"interleaved_thinking":false,"model_context_length":65536,"proactive_summarization_threshold":40960(...TRUNCATED) | false | null | 2026-07-30T09:27:23.505086Z | 15 | 0 | 3,368 | 32,671 | 1.1 | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [REDACTED] | Qwen/Qwen3.5-4B | 76 | 0 | 271,681 | 15,078 | 0 | 3 | false | 0 | 0 | 0 | {"enable_thinking":true,"extra_body":{"chat_template_kwargs":{"enable_thinking":true},"min_p":0.0,"r(...TRUNCATED) | [REDACTED]_base-thinking-canonical-repaired-20260730 | 2026-07-30T09:12:27.461742Z | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | 30 | make-mips-interpreter | [{"name":"test_outputs.py::test_vm_execution","status":"failed"},{"name":"test_outputs.py::test_fram(...TRUNCATED) | true | preserve-zero-tag-generations-without-inference | true | 76 | true | {"base_present":true,"contiguous":false,"continuation_indices":[1,4,30],"missing_continuation_indice(...TRUNCATED) | make-mips-interpreter__G8osS55 | 0 |
terminus-2 | terminal-bench/terminal-bench-2-1 | {"interleaved_thinking":false,"model_context_length":65536,"proactive_summarization_threshold":40960(...TRUNCATED) | false | null | 2026-07-30T10:20:28.785467Z | 10 | 0 | 1,818 | 32,672 | 1.1 | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [REDACTED] | Qwen/Qwen3.5-4B | 104 | 0 | 344,908 | 16,730 | 4 | 5 | false | 0 | 0 | 0 | {"enable_thinking":true,"extra_body":{"chat_template_kwargs":{"enable_thinking":true},"min_p":0.0,"r(...TRUNCATED) | [REDACTED]_base-thinking-canonical-repaired-20260730 | 2026-07-30T10:02:44.103665Z | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | 23 | path-tracing | [{"name":"test_outputs.py::test_image_c_exists","status":"passed"},{"name":"test_outputs.py::test_im(...TRUNCATED) | true | preserve-zero-tag-generations-without-inference | true | 104 | true | {"base_present":true,"contiguous":false,"continuation_indices":[1,2,3,23],"missing_continuation_indi(...TRUNCATED) | path-tracing__rRZbDtt | 0 |
terminus-2 | terminal-bench/terminal-bench-2-1 | {"interleaved_thinking":false,"model_context_length":65536,"proactive_summarization_threshold":40960(...TRUNCATED) | false | null | 2026-07-30T11:02:38.860270Z | 9 | 0 | 3,129 | 30,446 | 1.1 | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [REDACTED] | Qwen/Qwen3.5-4B | 54 | 0 | 381,004 | 38,143 | 2 | 3 | false | 0 | 0 | 0 | {"enable_thinking":true,"extra_body":{"chat_template_kwargs":{"enable_thinking":true},"min_p":0.0,"r(...TRUNCATED) | [REDACTED]_base-thinking-canonical-repaired-20260730 | 2026-07-30T10:36:15.438816Z | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | 20 | winning-avg-corewars | [{"name":"test_outputs.py::test_opponents_integrity","status":"passed"},{"name":"test_outputs.py::te(...TRUNCATED) | true | preserve-zero-tag-generations-without-inference | true | 54 | true | {"base_present":true,"contiguous":false,"continuation_indices":[1,2,20],"missing_continuation_indice(...TRUNCATED) | winning-avg-corewars__4FRMu9Q | 0 |
terminus-2 | terminal-bench/terminal-bench-2-1 | {"interleaved_thinking":false,"model_context_length":65536,"proactive_summarization_threshold":40960(...TRUNCATED) | false | {
"message": "Agent execution timed out after 900.0 seconds",
"type": "AgentTimeoutError"
} | 2026-07-30T10:29:08.838851Z | 0 | 0 | 243 | 1,681 | 1.1 | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [REDACTED] | Qwen/Qwen3.5-4B | 2 | 0 | 3,142 | 387 | 0 | 9 | true | 0 | 0 | 0 | {"enable_thinking":true,"extra_body":{"chat_template_kwargs":{"enable_thinking":true},"min_p":0.0,"r(...TRUNCATED) | [REDACTED]_base-thinking-canonical-repaired-20260730 | 2026-07-30T10:13:32.244909Z | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | 0 | adaptive-rejection-sampler | [{"name":"test_outputs.py::test_ars_function_exists","status":"failed"},{"name":"test_outputs.py::te(...TRUNCATED) | true | preserve-zero-tag-generations-without-inference | true | 2 | true | {
"base_present": true,
"contiguous": true,
"continuation_indices": [],
"missing_continuation_indices": []
} | adaptive-rejection-sampler__6E8VRgC | 0 |
terminus-2 | terminal-bench/terminal-bench-2-1 | {"interleaved_thinking":false,"model_context_length":65536,"proactive_summarization_threshold":40960(...TRUNCATED) | false | {
"message": "Agent execution timed out after 900.0 seconds",
"type": "AgentTimeoutError"
} | 2026-07-30T09:16:10.546393Z | 0 | 0 | 307 | 1,717 | 1.1 | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [REDACTED] | Qwen/Qwen3.5-4B | 2 | 0 | 3,177 | 472 | 0 | 9 | true | 0 | 0 | 0 | {"enable_thinking":true,"extra_body":{"chat_template_kwargs":{"enable_thinking":true},"min_p":0.0,"r(...TRUNCATED) | [REDACTED]_base-thinking-canonical-repaired-20260730 | 2026-07-30T09:00:33.975213Z | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | 0 | adaptive-rejection-sampler | [{"name":"test_outputs.py::test_ars_function_exists","status":"failed"},{"name":"test_outputs.py::te(...TRUNCATED) | true | preserve-zero-tag-generations-without-inference | true | 2 | true | {
"base_present": true,
"contiguous": true,
"continuation_indices": [],
"missing_continuation_indices": []
} | adaptive-rejection-sampler__8rp8Wmu | 0 |
terminus-2 | terminal-bench/terminal-bench-2-1 | {"interleaved_thinking":false,"model_context_length":65536,"proactive_summarization_threshold":40960(...TRUNCATED) | false | {
"message": "Agent execution timed out after 900.0 seconds",
"type": "AgentTimeoutError"
} | 2026-07-30T08:35:12.352666Z | 0 | 0 | 331 | 1,757 | 1.1 | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [REDACTED] | Qwen/Qwen3.5-4B | 2 | 0 | 3,218 | 495 | 0 | 9 | true | 0 | 0 | 0 | {"enable_thinking":true,"extra_body":{"chat_template_kwargs":{"enable_thinking":true},"min_p":0.0,"r(...TRUNCATED) | [REDACTED]_base-thinking-canonical-repaired-20260730 | 2026-07-30T08:19:30.807530Z | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | 0 | adaptive-rejection-sampler | [{"name":"test_outputs.py::test_ars_function_exists","status":"failed"},{"name":"test_outputs.py::te(...TRUNCATED) | true | preserve-zero-tag-generations-without-inference | true | 2 | true | {
"base_present": true,
"contiguous": true,
"continuation_indices": [],
"missing_continuation_indices": []
} | adaptive-rejection-sampler__RK9XwMp | 0 |
terminus-2 | terminal-bench/terminal-bench-2-1 | {"interleaved_thinking":false,"model_context_length":65536,"proactive_summarization_threshold":40960(...TRUNCATED) | false | {
"message": "Agent execution timed out after 900.0 seconds",
"type": "AgentTimeoutError"
} | 2026-07-30T11:07:57.534325Z | 0 | 0 | 4,465 | 20,170 | 1.1 | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [REDACTED] | Qwen/Qwen3.5-4B | 9 | 0 | 93,045 | 13,667 | 7 | 9 | true | 0 | 0 | 0 | {"enable_thinking":true,"extra_body":{"chat_template_kwargs":{"enable_thinking":true},"min_p":0.0,"r(...TRUNCATED) | [REDACTED]_base-thinking-canonical-repaired-20260730 | 2026-07-30T10:52:21.740059Z | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | 0 | adaptive-rejection-sampler | [{"name":"test_outputs.py::test_ars_function_exists","status":"passed"},{"name":"test_outputs.py::te(...TRUNCATED) | true | preserve-zero-tag-generations-without-inference | true | 9 | true | {
"base_present": true,
"contiguous": true,
"continuation_indices": [],
"missing_continuation_indices": []
} | adaptive-rejection-sampler__evceFhD | 0 |
terminus-2 | terminal-bench/terminal-bench-2-1 | {"interleaved_thinking":false,"model_context_length":65536,"proactive_summarization_threshold":40960(...TRUNCATED) | false | {
"message": "Agent execution timed out after 900.0 seconds",
"type": "AgentTimeoutError"
} | 2026-07-30T09:52:08.777511Z | 0 | 0 | 306 | 1,688 | 1.1 | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"content":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [REDACTED] | Qwen/Qwen3.5-4B | 2 | 0 | 3,150 | 442 | 0 | 9 | true | 0 | 0 | 0 | {"enable_thinking":true,"extra_body":{"chat_template_kwargs":{"enable_thinking":true},"min_p":0.0,"r(...TRUNCATED) | [REDACTED]_base-thinking-canonical-repaired-20260730 | 2026-07-30T09:36:34.464594Z | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | [{"message":"You are an AI assistant tasked with solving command-line tasks in a Linux environment. (...TRUNCATED) | 0 | adaptive-rejection-sampler | [{"name":"test_outputs.py::test_ars_function_exists","status":"failed"},{"name":"test_outputs.py::te(...TRUNCATED) | true | preserve-zero-tag-generations-without-inference | true | 2 | true | {
"base_present": true,
"contiguous": true,
"continuation_indices": [],
"missing_continuation_indices": []
} | adaptive-rejection-sampler__wm3HmF6 | 0 |
End of preview. Expand in Data Studio
[REDACTED] — Terminal-Bench 2.1
Canonical Terminal-Bench 2.1 evaluation of Qwen/Qwen3.5-4B through the served
model ID [REDACTED] with Terminus-2.
Result
- Recorded trials: 445
- Tasks / attempts: 89 × 5
- Errored trials scored as zero: 92
- Exception counts:
{"AgentTimeoutError": 89, "Timeout": 1, "VerifierTimeoutError": 2} - Agent timeouts / context-length events / output-cap events: 89 / 0 / 0
- Mean reward / Pass@1: 0.105618
- Pass@5: 0.191011
- Total input/output/cache tokens: 118111514 / 5571672 / 0
- Proactive summaries: 385
Protocol
Temperature 0.6, top-p
0.95, top-k omitted (disabled),
max_tokens=32768,
enable_thinking=true, max turns
30, concurrency
30.
- Trials / untagged / malformed assistant turns with incomplete thinking markup: 0 / 0 / 0
- Assistant-turn thinking-markup completeness: 1.000000
The server context length was 65536;
Terminus proactive summarization began at
40960 tokens.
Task-native TB2.1 timeouts were unchanged (timeout_multiplier=1.0). All errored trials remain in the denominator with reward
zero.
Viewer columns
messages and messages_without_thinking are aligned, strictly alternating
user/assistant conversations. Terminal observations are user messages.
steps and steps_without_thinking retain the detailed Harbor trajectory.
For thinking mode, Qwen's prompt-supplied opening <think> marker is
reconstructed while generated reasoning and </think> are preserved.
Zero-tag generations are preserved byte-for-byte in both views without
inference and are explicitly marked by thinking_markup_complete,
untagged_assistant_turn_count, and thinking_markup_policy.
trajectory_integrity reports which raw continuation segments were available;
missing segments are never fabricated.
Harbor's literal technical-difficulties fallback is marked with an empty
<think></think> block only in the normalized full-thinking view; its count is
recorded separately and the raw fallback text is retained.
Only verifier test names/statuses are published. Hidden verifier traces and source, authorization headers, API keys, and credential values are excluded or redacted.
- Downloads last month
- 34