{
"results": {
"aime24_nofigures": {
"alias": "aime24_nofigures",
"exact_match,none": 0.5666666666666667,
"exact_match_stderr,none": "N/A",
"extracted_answers,none": -1,
"extracted_answers_stderr,none": "N/A"
},
"aime25_nofigures": {
"alias": "aime25_nofigures",
"exact_match,none": 0.5333333333333333,
"exact_match_stderr,none": "N/A",
"extracted_answers,none": -1,
"extracted_answers_stderr,none": "N/A"
}
},
"group_subtasks": {
"aime24_nofigures": [],
"aime25_nofigures": []
},
"configs": {
"aime24_nofigures": {
"task": "aime24_nofigures",
"tag": [
"math_word_problems"
],
"dataset_path": "simplescaling/aime24_nofigures",
"dataset_name": "default",
"test_split": "train",
"process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n def _process_doc(doc: dict) -> dict:\n solution = doc.get(\"solution\", doc.get(\"orig_solution\", doc.get(\"orig_orig_solution\")))\n problem = doc.get(\"problem\", doc.get(\"question\"))\n answer = doc.get(\"answer\", doc.get(\"orig_answer\", doc.get(\"orig_orig_answer\")))\n if solution is None:\n print(\"Warning: No solution found; DOC:\", doc)\n out_doc = {\n \"problem\": problem,\n \"solution\": solution,\n \"answer\": answer,\n }\n if getattr(doc, \"few_shot\", None) is not None:\n out_doc[\"few_shot\"] = True\n return out_doc\n return dataset.map(_process_doc)\n",
"doc_to_text": "def doc_to_text(doc: dict) -> str:\n return QUERY_TEMPLATE.format(Question=doc.get(\"problem\", doc.get(\"question\")))\n",
"doc_to_target": "answer",
"process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n metrics = {\"exact_match\": None, \"extracted_answers\": []}\n # bp()\n # Multiple results -> we are measuring cov/maj etc\n if isinstance(results[0], list):\n results = results[0]\n n_res = len(results) # e.g. 64\n n_res_list = [2**i for i in range(1, int(n_res.bit_length()))] # e.g. [2, 4, 8, 16, 32, 64]\n metrics = {\n **metrics,\n \"exact_matches\": [],\n **{f\"cov@{n}\": -1 for n in n_res_list},\n **{f\"maj@{n}\": -1 for n in n_res_list},\n }\n\n if os.getenv(\"PROCESSOR\", \"\") == \"gpt-4o-mini\":\n sampler = ChatCompletionSampler(model=\"gpt-4o-mini\")\n else:\n print(f\"Unknown processor: {os.getenv('PROCESSOR')}; set 'PROCESSOR=gpt-4o-mini' and 'OPENAI_API_KEY=YOUR_KEY' for best results.\")\n sampler = None\n\n if isinstance(doc[\"answer\"], str) and doc[\"answer\"].isdigit():\n gt = str(int(doc[\"answer\"])) # 023 -> 23\n else:\n gt = str(doc[\"answer\"])\n split_tokens = [\"<|im_start|>answer\\n\", \"<|im_start|>\"]\n\n for i, a in enumerate(results, start=1):\n if split_tokens[0] in a:\n a = a.split(split_tokens[0])[-1]\n elif split_tokens[1] in a:\n a = a.split(split_tokens[1])[-1]\n if \"\\n\" in a:\n a = \"\\n\".join(a.split(\"\\n\")[1:])\n\n if (box := last_boxed_only_string(a)) is not None:\n a = remove_boxed(box)\n # re.DOTALL is key such that newlines are included e.g. if it does `Answer: Here is the solution:\\n\\n10`\n elif (matches := re.findall(ANSWER_PATTERN, a, re.DOTALL)) != []:\n a = matches[-1] # Get the last match\n\n # AIME answers are from 000 to 999 so often it is a digit anyways\n if (a.isdigit()) and (gt.isdigit()):\n a = str(int(a)) # 023 -> 23\n elif sampler is not None:\n options = [gt] + list(set(metrics[\"extracted_answers\"]) - {gt})\n if len(options) > 7:\n # Could switch back to exact returning like in AIME in that case\n # Problem with exact returning is that it sometimes messes up small things like a dollar sign\n print(\"Warning: Lots of options which may harm indexing performance:\", options) \n # This ensures that if doc['answer'] is \\text{Evelyn} it is represented as such and not \\\\text{Evelyn}\n options_str = \"[\" + \", \".join([\"'\" + str(o) + \"'\" for o in options]) + \"]\"\n # a = extract_answer(sampler, options, a)\n idx = extract_answer_idx(sampler, options_str, a)\n if idx != \"-1\":\n if idx.isdigit():\n idx = int(idx) - 1\n if len(options) > idx >= 0:\n a = options[idx]\n else:\n print(\"Warning: Index out of bounds; leaving answer unchanged\\n\", a, \"\\noptions\", options_str, \"\\ndoc['answer']\", gt, \"\\nidx\", idx)\n else:\n print(\"Warning: Processing did not produce integer index\\na\", a, \"\\noptions\", options_str, \"\\ndoc['answer']\", gt, \"\\nidx\", idx)\n else:\n pass # TODO: Maybe add back legacy processing\n\n metrics[\"extracted_answers\"].append(a)\n a = int(a == gt)\n if not(a): # Optional logging\n print(\"Marked incorrect\\na \" + metrics[\"extracted_answers\"][-1] + \"\\ndoc['answer'] \" + gt)\n if i == 1:\n metrics[\"exact_match\"] = a\n if \"exact_matches\" in metrics:\n metrics[\"exact_matches\"].append(a)\n elif i > 1:\n metrics[\"exact_matches\"].append(a)\n if i in n_res_list:\n metrics[f\"cov@{i}\"] = int(1 in metrics[\"exact_matches\"])\n metrics[f\"maj@{i}\"] = int(gt == Counter(metrics[\"extracted_answers\"]).most_common(1)[0][0])\n\n return metrics\n",
"description": "",
"target_delimiter": " ",
"fewshot_delimiter": "\n\n",
"num_fewshot": 0,
"metric_list": [
{
"metric": "exact_match",
"aggregation": "mean",
"higher_is_better": true
},
{
"metric": "extracted_answers",
"aggregation": "bypass",
"higher_is_better": true
}
],
"output_type": "generate_until",
"generation_kwargs": {
"until": [],
"do_sample": false,
"temperature": 0.0,
"max_gen_toks": 32768,
"max_tokens_thinking": "auto",
"thinking_n_ignore": 2,
"thinking_n_ignore_str": "wait..."
},
"repeats": 1,
"should_decontaminate": false,
"metadata": {
"version": 1.0
}
},
"aime25_nofigures": {
"task": "aime25_nofigures",
"tag": [
"math_word_problems"
],
"dataset_path": "TIGER-Lab/AIME25",
"dataset_name": "default",
"test_split": "train",
"process_docs": "def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:\n def _process_doc(doc: dict) -> dict:\n solution = doc.get(\"solution\", doc.get(\"orig_solution\", doc.get(\"orig_orig_solution\")))\n problem = doc.get(\"problem\", doc.get(\"question\"))\n answer = doc.get(\"answer\", doc.get(\"orig_answer\", doc.get(\"orig_orig_answer\")))\n if solution is None:\n print(\"Warning: No solution found; DOC:\", doc)\n out_doc = {\n \"problem\": problem,\n \"solution\": solution,\n \"answer\": answer,\n }\n if getattr(doc, \"few_shot\", None) is not None:\n out_doc[\"few_shot\"] = True\n return out_doc\n return dataset.map(_process_doc)\n",
"doc_to_text": "def doc_to_text(doc: dict) -> str:\n return QUERY_TEMPLATE.format(Question=doc.get(\"problem\", doc.get(\"question\")))\n",
"doc_to_target": "answer",
"process_results": "def process_results(doc: dict, results: List[str]) -> Dict[str, int]:\n metrics = {\"exact_match\": None, \"extracted_answers\": []}\n # bp()\n # Multiple results -> we are measuring cov/maj etc\n if isinstance(results[0], list):\n results = results[0]\n n_res = len(results) # e.g. 64\n n_res_list = [2**i for i in range(1, int(n_res.bit_length()))] # e.g. [2, 4, 8, 16, 32, 64]\n metrics = {\n **metrics,\n \"exact_matches\": [],\n **{f\"cov@{n}\": -1 for n in n_res_list},\n **{f\"maj@{n}\": -1 for n in n_res_list},\n }\n\n if os.getenv(\"PROCESSOR\", \"\") == \"gpt-4o-mini\":\n sampler = ChatCompletionSampler(model=\"gpt-4o-mini\")\n else:\n print(f\"Unknown processor: {os.getenv('PROCESSOR')}; set 'PROCESSOR=gpt-4o-mini' and 'OPENAI_API_KEY=YOUR_KEY' for best results.\")\n sampler = None\n\n if isinstance(doc[\"answer\"], str) and doc[\"answer\"].isdigit():\n gt = str(int(doc[\"answer\"])) # 023 -> 23\n else:\n gt = str(doc[\"answer\"])\n split_tokens = [\"<|im_start|>answer\\n\", \"<|im_start|>\"]\n\n for i, a in enumerate(results, start=1):\n if split_tokens[0] in a:\n a = a.split(split_tokens[0])[-1]\n elif split_tokens[1] in a:\n a = a.split(split_tokens[1])[-1]\n if \"\\n\" in a:\n a = \"\\n\".join(a.split(\"\\n\")[1:])\n\n if (box := last_boxed_only_string(a)) is not None:\n a = remove_boxed(box)\n # re.DOTALL is key such that newlines are included e.g. if it does `Answer: Here is the solution:\\n\\n10`\n elif (matches := re.findall(ANSWER_PATTERN, a, re.DOTALL)) != []:\n a = matches[-1] # Get the last match\n\n # AIME answers are from 000 to 999 so often it is a digit anyways\n if (a.isdigit()) and (gt.isdigit()):\n a = str(int(a)) # 023 -> 23\n elif sampler is not None:\n options = [gt] + list(set(metrics[\"extracted_answers\"]) - {gt})\n if len(options) > 7:\n # Could switch back to exact returning like in AIME in that case\n # Problem with exact returning is that it sometimes messes up small things like a dollar sign\n print(\"Warning: Lots of options which may harm indexing performance:\", options) \n # This ensures that if doc['answer'] is \\text{Evelyn} it is represented as such and not \\\\text{Evelyn}\n options_str = \"[\" + \", \".join([\"'\" + str(o) + \"'\" for o in options]) + \"]\"\n # a = extract_answer(sampler, options, a)\n idx = extract_answer_idx(sampler, options_str, a)\n if idx != \"-1\":\n if idx.isdigit():\n idx = int(idx) - 1\n if len(options) > idx >= 0:\n a = options[idx]\n else:\n print(\"Warning: Index out of bounds; leaving answer unchanged\\n\", a, \"\\noptions\", options_str, \"\\ndoc['answer']\", gt, \"\\nidx\", idx)\n else:\n print(\"Warning: Processing did not produce integer index\\na\", a, \"\\noptions\", options_str, \"\\ndoc['answer']\", gt, \"\\nidx\", idx)\n else:\n pass # TODO: Maybe add back legacy processing\n\n metrics[\"extracted_answers\"].append(a)\n a = int(a == gt)\n if not(a): # Optional logging\n print(\"Marked incorrect\\na \" + metrics[\"extracted_answers\"][-1] + \"\\ndoc['answer'] \" + gt)\n if i == 1:\n metrics[\"exact_match\"] = a\n if \"exact_matches\" in metrics:\n metrics[\"exact_matches\"].append(a)\n elif i > 1:\n metrics[\"exact_matches\"].append(a)\n if i in n_res_list:\n metrics[f\"cov@{i}\"] = int(1 in metrics[\"exact_matches\"])\n metrics[f\"maj@{i}\"] = int(gt == Counter(metrics[\"extracted_answers\"]).most_common(1)[0][0])\n\n return metrics\n",
"description": "",
"target_delimiter": " ",
"fewshot_delimiter": "\n\n",
"num_fewshot": 0,
"metric_list": [
{
"metric": "exact_match",
"aggregation": "mean",
"higher_is_better": true
},
{
"metric": "extracted_answers",
"aggregation": "bypass",
"higher_is_better": true
}
],
"output_type": "generate_until",
"generation_kwargs": {
"until": [],
"do_sample": false,
"temperature": 0.0,
"max_gen_toks": 32768,
"max_tokens_thinking": "auto",
"thinking_n_ignore": 2,
"thinking_n_ignore_str": "wait..."
},
"repeats": 1,
"should_decontaminate": false,
"metadata": {
"version": 1.0
}
}
},
"versions": {
"aime24_nofigures": 1.0,
"aime25_nofigures": 1.0
},
"n-shot": {
"aime24_nofigures": 0,
"aime25_nofigures": 0
},
"higher_is_better": {
"aime24_nofigures": {
"exact_match": true,
"extracted_answers": true
},
"aime25_nofigures": {
"exact_match": true,
"extracted_answers": true
}
},
"n-samples": {
"aime25_nofigures": {
"original": 15,
"effective": 15
},
"aime24_nofigures": {
"original": 30,
"effective": 30
}
},
"config": {
"model": "vllm",
"model_args": "pretrained=simplescaling/s1.1-32B,dtype=float32,tensor_parallel_size=8",
"batch_size": "auto",
"batch_sizes": [],
"device": null,
"use_cache": null,
"limit": null,
"bootstrap_iters": 0,
"gen_kwargs": {
"max_gen_toks": 32768,
"max_tokens_thinking": "auto",
"thinking_n_ignore": 2,
"thinking_n_ignore_str": "wait..."
},
"random_seed": 0,
"numpy_seed": 1234,
"torch_seed": 1234,
"fewshot_seed": 1234
},
"git_hash": "a465d7f",
"date": 1739478144.3046002,
"pretty_env_info": "PyTorch version: 2.5.1+cu124\nIs debug build: False\nCUDA used to build PyTorch: 12.4\nROCM used to build PyTorch: N/A\n\nOS: Ubuntu 22.04.3 LTS (x86_64)\nGCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0\nClang version: Could not collect\nCMake version: Could not collect\nLibc version: glibc-2.35\n\nPython version: 3.10.12 (main, Jun 11 2023, 05:26:28) [GCC 11.4.0] (64-bit runtime)\nPython platform: Linux-6.5.0-45-generic-x86_64-with-glibc2.35\nIs CUDA available: True\nCUDA runtime version: 11.8.89\nCUDA_MODULE_LOADING set to: LAZY\nGPU models and configuration: \nGPU 0: NVIDIA A40\nGPU 1: NVIDIA A40\nGPU 2: NVIDIA A40\nGPU 3: NVIDIA A40\nGPU 4: NVIDIA A40\nGPU 5: NVIDIA A40\nGPU 6: NVIDIA A40\nGPU 7: NVIDIA A40\n\nNvidia driver version: 550.127.05\ncuDNN version: Could not collect\nHIP runtime version: N/A\nMIOpen runtime version: N/A\nIs XNNPACK available: True\n\nCPU:\nArchitecture: x86_64\nCPU op-mode(s): 32-bit, 64-bit\nAddress sizes: 52 bits physical, 57 bits virtual\nByte Order: Little Endian\nCPU(s): 96\nOn-line CPU(s) list: 0-95\nVendor ID: GenuineIntel\nModel name: Intel(R) Xeon(R) Gold 6342 CPU @ 2.80GHz\nCPU family: 6\nModel: 106\nThread(s) per core: 2\nCore(s) per socket: 24\nSocket(s): 2\nStepping: 6\nCPU max MHz: 3500.0000\nCPU min MHz: 800.0000\nBogoMIPS: 5600.00\nFlags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 invpcid_single ssbd mba ibrs ibpb stibp ibrs_enhanced tpr_shadow flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb intel_pt avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local split_lock_detect wbnoinvd dtherm ida arat pln pts vnmi avx512vbmi umip pku ospke avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg tme avx512_vpopcntdq la57 rdpid fsrm md_clear pconfig flush_l1d arch_capabilities\nVirtualization: VT-x\nL1d cache: 2.3 MiB (48 instances)\nL1i cache: 1.5 MiB (48 instances)\nL2 cache: 60 MiB (48 instances)\nL3 cache: 72 MiB (2 instances)\nNUMA node(s): 2\nNUMA node0 CPU(s): 0-23,48-71\nNUMA node1 CPU(s): 24-47,72-95\nVulnerability Gather data sampling: Mitigation; Microcode\nVulnerability Itlb multihit: Not affected\nVulnerability L1tf: Not affected\nVulnerability Mds: Not affected\nVulnerability Meltdown: Not affected\nVulnerability Mmio stale data: Mitigation; Clear CPU buffers; SMT vulnerable\nVulnerability Retbleed: Not affected\nVulnerability Spec rstack overflow: Not affected\nVulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl\nVulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2: Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI Syscall hardening, KVM SW loop\nVulnerability Srbds: Not affected\nVulnerability Tsx async abort: Not affected\n\nVersions of relevant libraries:\n[pip3] numpy==1.26.4\n[pip3] torch==2.5.1\n[pip3] torchaudio==2.5.1\n[pip3] torchvision==0.20.1\n[pip3] triton==3.1.0\n[conda] Could not collect",
"transformers_version": "4.48.3",
"upper_git_hash": null,
"tokenizer_pad_token": [
"<|endoftext|>",
"151643"
],
"tokenizer_eos_token": [
"<|im_end|>",
"151645"
],
"tokenizer_bos_token": [
null,
"None"
],
"eot_token_id": 151645,
"max_length": 32768,
"task_hashes": {
"aime25_nofigures": "62c256d557633e4054a8cd84e22eb1971f82e7c4cee36cb508cf37362cf7f67a",
"aime24_nofigures": "3eb5fb976b3f4dea4e4e2a2caf5efa2cfea98aa3ae68cd0f3bfa8a3f197b0e2d"
},
"model_source": "vllm",
"model_name": "simplescaling/s1.1-32B",
"model_name_sanitized": "simplescaling__s1.1-32B",
"system_instruction": null,
"system_instruction_sha": null,
"fewshot_as_multiturn": false,
"chat_template": "{%- if tools %}\n {{- '<|im_start|>system\\n' }}\n {%- if messages[0]['role'] == 'system' %}\n {{- messages[0]['content'] }}\n {%- else %}\n {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}\n {%- endif %}\n {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within XML tags:\\n\" }}\n {%- for tool in tools %}\n {{- \"\\n\" }}\n {{- tool | tojson }}\n {%- endfor %}\n {{- \"\\n\\n\\nFor each function call, return a json object with function name and arguments within XML tags:\\n\\n{\\\"name\\\": , \\\"arguments\\\": }\\n<|im_end|>\\n\" }}\n{%- else %}\n {%- if messages[0]['role'] == 'system' %}\n {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n {%- else %}\n {{- '<|im_start|>system\\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\\n' }}\n {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n {%- elif message.role == \"assistant\" %}\n {{- '<|im_start|>' + message.role }}\n {%- if message.content %}\n {{- '\\n' + message.content }}\n {%- endif %}\n {%- for tool_call in message.tool_calls %}\n {%- if tool_call.function is defined %}\n {%- set tool_call = tool_call.function %}\n {%- endif %}\n {{- '\\n\\n{\"name\": \"' }}\n {{- tool_call.name }}\n {{- '\", \"arguments\": ' }}\n {{- tool_call.arguments | tojson }}\n {{- '}\\n' }}\n {%- endfor %}\n {{- '<|im_end|>\\n' }}\n {%- elif message.role == \"tool\" %}\n {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n {{- '<|im_start|>user' }}\n {%- endif %}\n {{- '\\n\\n' }}\n {{- message.content }}\n {{- '\\n' }}\n {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n {{- '<|im_end|>\\n' }}\n {%- endif %}\n {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
"chat_template_sha": "cd8e9439f0570856fd70470bf8889ebd8b5d1107207f67a5efb46e342330527f",
"start_time": 8663445.301817764,
"end_time": 8669003.429823114,
"total_evaluation_time_seconds": "5558.128005350009"
}