run stringclasses 6
values | session stringclasses 590
values | arm stringclasses 3
values | harness stringclasses 5
values | task stringclasses 88
values | trial stringclasses 584
values | step int64 1 153 | ts stringlengths 29 29 | p_hard float64 0 1 ⌀ | tier stringclasses 3
values | reason stringclasses 3
values | state stringlengths 395 9.59k |
|---|---|---|---|---|---|---|---|---|---|---|---|
test-lite | jah-auto-t75/llm-inference-batching-scheduler__LSuLjLc | jah-auto-t75 | pi | llm-inference-batching-scheduler | llm-inference-batching-scheduler__LSuLjLc | 20 | 2026-10-15T00:16:15.023+00:00 | 0.33 | fast | routine | TASK: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to operate on fixed‑sized ten... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__LSuLjLc | jah-auto-t75 | pi | llm-inference-batching-scheduler | llm-inference-batching-scheduler__LSuLjLc | 21 | 2026-10-15T00:16:16.220+00:00 | 0.24 | fast | routine | TASK: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to operate on fixed‑sized ten... |
test-lite | jah-auto-t75/gpt2-codegolf__Go76jG5 | jah-auto-t75 | t2 | gpt2-codegolf | gpt2-codegolf__Go76jG5 | 7 | 2026-10-15T00:16:16.780+00:00 | 0.6 | fast | routine | TASK: You are an AI assistant tasked with solving command-line tasks in a Linux environment. You will be given a task description and the output from previously executed commands. Your goal is to solve the task by providing batches of shell commands. Format your response as JSON with the following structure: { "analysi... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__LSuLjLc | jah-auto-t75 | pi | llm-inference-batching-scheduler | llm-inference-batching-scheduler__LSuLjLc | 22 | 2026-10-15T00:16:17.654+00:00 | 0.1 | fast | routine | TASK: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to operate on fixed‑sized ten... |
test-lite | jah-auto-t75/gpt2-codegolf__wbNvMPD | jah-auto-t75 | pi | gpt2-codegolf | gpt2-codegolf__wbNvMPD | 23 | 2026-10-15T00:16:18.311+00:00 | 0.94 | strong | hard | TASK: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I will run it /app/a.out gp... |
test-lite | jah-auto-t75/gpt2-codegolf__Go76jG5 | jah-auto-t75 | t2 | gpt2-codegolf | gpt2-codegolf__Go76jG5 | 8 | 2026-10-15T00:16:18.649+00:00 | 0.65 | fast | routine | TASK: You are an AI assistant tasked with solving command-line tasks in a Linux environment. You will be given a task description and the output from previously executed commands. Your goal is to solve the task by providing batches of shell commands. Format your response as JSON with the following structure: { "analysi... |
test-lite | jah-auto-t75/gpt2-codegolf__Go76jG5 | jah-auto-t75 | t2 | gpt2-codegolf | gpt2-codegolf__Go76jG5 | 9 | 2026-10-15T00:16:20.871+00:00 | 0.86 | strong | hard | TASK: You are an AI assistant tasked with solving command-line tasks in a Linux environment. You will be given a task description and the output from previously executed commands. Your goal is to solve the task by providing batches of shell commands. Format your response as JSON with the following structure: { "analysi... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__NzkUYEV | jah-auto-t75 | t2 | llm-inference-batching-scheduler | llm-inference-batching-scheduler__NzkUYEV | 6 | 2026-10-15T00:16:22.632+00:00 | 0.84 | strong | hard | TASK: You are an AI assistant tasked with solving command-line tasks in a Linux environment. You will be given a task description and the output from previously executed commands. Your goal is to solve the task by providing batches of shell commands. Format your response as JSON with the following structure: { "analysi... |
test-lite | jah-auto-t75/gpt2-codegolf__D5XWnZ6 | jah-auto-t75 | mswe | gpt2-codegolf | gpt2-codegolf__D5XWnZ6 | 15 | 2026-10-15T00:16:22.852+00:00 | 0.71 | fast | routine | TASK: Please solve this issue: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I ... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 15 | 2026-10-15T00:16:24.824+00:00 | 0.74 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/gpt2-codegolf__D5XWnZ6 | jah-auto-t75 | mswe | gpt2-codegolf | gpt2-codegolf__D5XWnZ6 | 16 | 2026-10-15T00:16:25.278+00:00 | 0.83 | strong | hard | TASK: Please solve this issue: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I ... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__8CB8D38 | jah-auto-t75 | mswe | llm-inference-batching-scheduler | llm-inference-batching-scheduler__8CB8D38 | 12 | 2026-10-15T00:16:26.516+00:00 | 0.85 | strong | hard | TASK: Please solve this issue: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to o... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 16 | 2026-10-15T00:16:27.353+00:00 | 0.51 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 17 | 2026-10-15T00:16:30.004+00:00 | 0.67 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 18 | 2026-10-15T00:16:32.148+00:00 | 0.62 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/break-filter-js-from-html__XaV6rcT | jah-auto-t75 | t2 | break-filter-js-from-html | break-filter-js-from-html__XaV6rcT | 10 | 2026-10-15T00:16:33.259+00:00 | 0.95 | strong | hard | TASK: You are an AI assistant tasked with solving command-line tasks in a Linux environment. You will be given a task description and the output from previously executed commands. Your goal is to solve the task by providing batches of shell commands. Format your response as JSON with the following structure: { "analysi... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 19 | 2026-10-15T00:16:34.339+00:00 | 0.47 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__LSuLjLc | jah-auto-t75 | pi | llm-inference-batching-scheduler | llm-inference-batching-scheduler__LSuLjLc | 23 | 2026-10-15T00:16:35.267+00:00 | 0.09 | fast | routine | TASK: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to operate on fixed‑sized ten... |
test-lite | jah-auto-t75/gpt2-codegolf__ij4jD4w | jah-auto-t75 | dsh | gpt2-codegolf | gpt2-codegolf__ij4jD4w | 28 | 2026-10-15T00:16:35.777+00:00 | 0.59 | fast | routine | TASK: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I will run it /app/a.out gp... |
test-lite | jah-auto-t75/gpt2-codegolf__wbNvMPD | jah-auto-t75 | pi | gpt2-codegolf | gpt2-codegolf__wbNvMPD | 24 | 2026-10-15T00:16:35.804+00:00 | 0.87 | strong | hard | TASK: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I will run it /app/a.out gp... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 20 | 2026-10-15T00:16:37.059+00:00 | 0.55 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__vzKDdmM | jah-auto-t75 | dsh | llm-inference-batching-scheduler | llm-inference-batching-scheduler__vzKDdmM | 12 | 2026-10-15T00:16:37.916+00:00 | 0.83 | strong | hard | TASK: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to operate on fixed‑sized ten... |
test-lite | jah-auto-t75/gpt2-codegolf__ij4jD4w | jah-auto-t75 | dsh | gpt2-codegolf | gpt2-codegolf__ij4jD4w | 29 | 2026-10-15T00:16:39.574+00:00 | 0.4 | fast | routine | TASK: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I will run it /app/a.out gp... |
test-lite | jah-auto-t75/reshard-c4-data__yaojNSw | jah-auto-t75 | pi | reshard-c4-data | reshard-c4-data__yaojNSw | 25 | 2026-10-15T00:16:43.326+00:00 | 0.11 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 21 | 2026-10-15T00:16:43.884+00:00 | 0.49 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__8CB8D38 | jah-auto-t75 | mswe | llm-inference-batching-scheduler | llm-inference-batching-scheduler__8CB8D38 | 13 | 2026-10-15T00:16:46.813+00:00 | 0.98 | strong | hard | TASK: Please solve this issue: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to o... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 22 | 2026-10-15T00:16:48.579+00:00 | 0.22 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 23 | 2026-10-15T00:16:50.974+00:00 | 0.99 | strong | hard | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/reshard-c4-data__yaojNSw | jah-auto-t75 | pi | reshard-c4-data | reshard-c4-data__yaojNSw | 26 | 2026-10-15T00:16:51.977+00:00 | 0.38 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/reshard-c4-data__yaojNSw | jah-auto-t75 | pi | reshard-c4-data | reshard-c4-data__yaojNSw | 27 | 2026-10-15T00:16:55.995+00:00 | 0.37 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__LSuLjLc | jah-auto-t75 | pi | llm-inference-batching-scheduler | llm-inference-batching-scheduler__LSuLjLc | 24 | 2026-10-15T00:16:57.283+00:00 | 0.03 | fast | routine | TASK: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to operate on fixed‑sized ten... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__vzKDdmM | jah-auto-t75 | dsh | llm-inference-batching-scheduler | llm-inference-batching-scheduler__vzKDdmM | 13 | 2026-10-15T00:16:59.739+00:00 | 0.56 | fast | routine | TASK: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to operate on fixed‑sized ten... |
test-lite | jah-auto-t75/write-compressor__5cP4Man | jah-auto-t75 | mswe | write-compressor | write-compressor__5cP4Man | 1 | 2026-10-15T00:17:01.097+00:00 | 0.48 | fast | routine | TASK: Please solve this issue: I have a decompressor in /app/decomp.c. It reads compressed data from stdin and writes the decompressed data to stdout. I also have a file /app/data.txt that has a bunch of text. Write me data.comp that's compressed such that running cat data.comp | /app/decomp gives exactly data.txt. You... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__vzKDdmM | jah-auto-t75 | dsh | llm-inference-batching-scheduler | llm-inference-batching-scheduler__vzKDdmM | 14 | 2026-10-15T00:17:01.765+00:00 | 0.59 | fast | routine | TASK: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to operate on fixed‑sized ten... |
test-lite | jah-auto-t75/write-compressor__5cP4Man | jah-auto-t75 | mswe | write-compressor | write-compressor__5cP4Man | 2 | 2026-10-15T00:17:02.475+00:00 | 0.77 | strong | hard | TASK: Please solve this issue: I have a decompressor in /app/decomp.c. It reads compressed data from stdin and writes the decompressed data to stdout. I also have a file /app/data.txt that has a bunch of text. Write me data.comp that's compressed such that running cat data.comp | /app/decomp gives exactly data.txt. You... |
test-lite | jah-auto-t75/break-filter-js-from-html__XaV6rcT | jah-auto-t75 | t2 | break-filter-js-from-html | break-filter-js-from-html__XaV6rcT | 11 | 2026-10-15T00:17:02.886+00:00 | 0.95 | strong | hard | TASK: You are an AI assistant tasked with solving command-line tasks in a Linux environment. You will be given a task description and the output from previously executed commands. Your goal is to solve the task by providing batches of shell commands. Format your response as JSON with the following structure: { "analysi... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__vzKDdmM | jah-auto-t75 | dsh | llm-inference-batching-scheduler | llm-inference-batching-scheduler__vzKDdmM | 15 | 2026-10-15T00:17:04.416+00:00 | 0.53 | fast | routine | TASK: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to operate on fixed‑sized ten... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__vzKDdmM | jah-auto-t75 | dsh | llm-inference-batching-scheduler | llm-inference-batching-scheduler__vzKDdmM | 16 | 2026-10-15T00:17:07.022+00:00 | 0.89 | strong | hard | TASK: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to operate on fixed‑sized ten... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__8CB8D38 | jah-auto-t75 | mswe | llm-inference-batching-scheduler | llm-inference-batching-scheduler__8CB8D38 | 14 | 2026-10-15T00:17:10.496+00:00 | 0.83 | strong | hard | TASK: Please solve this issue: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to o... |
test-lite | jah-auto-t75/reshard-c4-data__Fjsg4Qb | jah-auto-t75 | t2 | reshard-c4-data | reshard-c4-data__Fjsg4Qb | 7 | 2026-10-15T00:17:11.744+00:00 | 0.78 | strong | hard | TASK: You are an AI assistant tasked with solving command-line tasks in a Linux environment. You will be given a task description and the output from previously executed commands. Your goal is to solve the task by providing batches of shell commands. Format your response as JSON with the following structure: { "analysi... |
test-lite | jah-auto-t75/gpt2-codegolf__D5XWnZ6 | jah-auto-t75 | mswe | gpt2-codegolf | gpt2-codegolf__D5XWnZ6 | 17 | 2026-10-15T00:17:18.079+00:00 | 0.79 | strong | hard | TASK: Please solve this issue: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I ... |
test-lite | jah-auto-t75/gpt2-codegolf__wbNvMPD | jah-auto-t75 | pi | gpt2-codegolf | gpt2-codegolf__wbNvMPD | 25 | 2026-10-15T00:17:19.594+00:00 | 0.87 | strong | hard | TASK: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I will run it /app/a.out gp... |
test-lite | jah-auto-t75/gpt2-codegolf__D5XWnZ6 | jah-auto-t75 | mswe | gpt2-codegolf | gpt2-codegolf__D5XWnZ6 | 18 | 2026-10-15T00:17:25.860+00:00 | 0.4 | fast | routine | TASK: Please solve this issue: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I ... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__8CB8D38 | jah-auto-t75 | mswe | llm-inference-batching-scheduler | llm-inference-batching-scheduler__8CB8D38 | 15 | 2026-10-15T00:17:28.681+00:00 | 0.5 | fast | routine | TASK: Please solve this issue: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to o... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__8CB8D38 | jah-auto-t75 | mswe | llm-inference-batching-scheduler | llm-inference-batching-scheduler__8CB8D38 | 16 | 2026-10-15T00:17:30.916+00:00 | 0.23 | fast | routine | TASK: Please solve this issue: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to o... |
test-lite | jah-auto-t75/break-filter-js-from-html__XaV6rcT | jah-auto-t75 | t2 | break-filter-js-from-html | break-filter-js-from-html__XaV6rcT | 12 | 2026-10-15T00:17:32.856+00:00 | 0.97 | strong | hard | TASK: You are an AI assistant tasked with solving command-line tasks in a Linux environment. You will be given a task description and the output from previously executed commands. Your goal is to solve the task by providing batches of shell commands. Format your response as JSON with the following structure: { "analysi... |
test-lite | jah-auto-t75/gpt2-codegolf__D5XWnZ6 | jah-auto-t75 | mswe | gpt2-codegolf | gpt2-codegolf__D5XWnZ6 | 19 | 2026-10-15T00:17:34.076+00:00 | 0.75 | strong | hard | TASK: Please solve this issue: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I ... |
test-lite | jah-auto-t75/write-compressor__5cP4Man | jah-auto-t75 | mswe | write-compressor | write-compressor__5cP4Man | 3 | 2026-10-15T00:17:38.138+00:00 | 0.78 | strong | hard | TASK: Please solve this issue: I have a decompressor in /app/decomp.c. It reads compressed data from stdin and writes the decompressed data to stdout. I also have a file /app/data.txt that has a bunch of text. Write me data.comp that's compressed such that running cat data.comp | /app/decomp gives exactly data.txt. You... |
test-lite | jah-auto-t75/gpt2-codegolf__ij4jD4w | jah-auto-t75 | dsh | gpt2-codegolf | gpt2-codegolf__ij4jD4w | 30 | 2026-10-15T00:17:45.320+00:00 | 0.68 | fast | routine | TASK: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I will run it /app/a.out gp... |
test-lite | jah-auto-t75/write-compressor__5cP4Man | jah-auto-t75 | mswe | write-compressor | write-compressor__5cP4Man | 4 | 2026-10-15T00:17:45.847+00:00 | 0.69 | fast | routine | TASK: Please solve this issue: I have a decompressor in /app/decomp.c. It reads compressed data from stdin and writes the decompressed data to stdout. I also have a file /app/data.txt that has a bunch of text. Write me data.comp that's compressed such that running cat data.comp | /app/decomp gives exactly data.txt. You... |
test-lite | jah-auto-t75/write-compressor__5cP4Man | jah-auto-t75 | mswe | write-compressor | write-compressor__5cP4Man | 5 | 2026-10-15T00:17:47.705+00:00 | 0.96 | strong | hard | TASK: Please solve this issue: I have a decompressor in /app/decomp.c. It reads compressed data from stdin and writes the decompressed data to stdout. I also have a file /app/data.txt that has a bunch of text. Write me data.comp that's compressed such that running cat data.comp | /app/decomp gives exactly data.txt. You... |
test-lite | jah-auto-t75/write-compressor__MqbbAej | jah-auto-t75 | pi | write-compressor | write-compressor__MqbbAej | 10 | 2026-10-15T00:17:48.982+00:00 | 0.83 | strong | hard | TASK: I have a decompressor in /app/decomp.c. It reads compressed data from stdin and writes the decompressed data to stdout. I also have a file /app/data.txt that has a bunch of text. Write me data.comp that's compressed such that running cat data.comp | /app/decomp gives exactly data.txt. You can generate data.comp a... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 24 | 2026-10-15T00:17:50.850+00:00 | 0.25 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 25 | 2026-10-15T00:17:52.228+00:00 | 0.55 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/reshard-c4-data__Fjsg4Qb | jah-auto-t75 | t2 | reshard-c4-data | reshard-c4-data__Fjsg4Qb | 8 | 2026-10-15T00:17:53.564+00:00 | 0.48 | fast | routine | TASK: You are an AI assistant tasked with solving command-line tasks in a Linux environment. You will be given a task description and the output from previously executed commands. Your goal is to solve the task by providing batches of shell commands. Format your response as JSON with the following structure: { "analysi... |
test-lite | jah-auto-t75/break-filter-js-from-html__XaV6rcT | jah-auto-t75 | t2 | break-filter-js-from-html | break-filter-js-from-html__XaV6rcT | 13 | 2026-10-15T00:17:54.060+00:00 | 0.96 | strong | hard | TASK: You are an AI assistant tasked with solving command-line tasks in a Linux environment. You will be given a task description and the output from previously executed commands. Your goal is to solve the task by providing batches of shell commands. Format your response as JSON with the following structure: { "analysi... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 26 | 2026-10-15T00:17:54.352+00:00 | 0.92 | strong | hard | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/write-compressor__5cP4Man | jah-auto-t75 | mswe | write-compressor | write-compressor__5cP4Man | 6 | 2026-10-15T00:17:54.975+00:00 | 0.92 | strong | hard | TASK: Please solve this issue: I have a decompressor in /app/decomp.c. It reads compressed data from stdin and writes the decompressed data to stdout. I also have a file /app/data.txt that has a bunch of text. Write me data.comp that's compressed such that running cat data.comp | /app/decomp gives exactly data.txt. You... |
test-lite | jah-auto-t75/reshard-c4-data__Fjsg4Qb | jah-auto-t75 | t2 | reshard-c4-data | reshard-c4-data__Fjsg4Qb | 9 | 2026-10-15T00:17:55.280+00:00 | 0.55 | fast | routine | TASK: You are an AI assistant tasked with solving command-line tasks in a Linux environment. You will be given a task description and the output from previously executed commands. Your goal is to solve the task by providing batches of shell commands. Format your response as JSON with the following structure: { "analysi... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__vzKDdmM | jah-auto-t75 | dsh | llm-inference-batching-scheduler | llm-inference-batching-scheduler__vzKDdmM | 17 | 2026-10-15T00:17:56.839+00:00 | 0.67 | fast | routine | TASK: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to operate on fixed‑sized ten... |
test-lite | jah-auto-t75/write-compressor__5cP4Man | jah-auto-t75 | mswe | write-compressor | write-compressor__5cP4Man | 7 | 2026-10-15T00:17:59.784+00:00 | 0.94 | strong | hard | TASK: Please solve this issue: I have a decompressor in /app/decomp.c. It reads compressed data from stdin and writes the decompressed data to stdout. I also have a file /app/data.txt that has a bunch of text. Write me data.comp that's compressed such that running cat data.comp | /app/decomp gives exactly data.txt. You... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__vzKDdmM | jah-auto-t75 | dsh | llm-inference-batching-scheduler | llm-inference-batching-scheduler__vzKDdmM | 18 | 2026-10-15T00:18:01.025+00:00 | 0.92 | strong | hard | TASK: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to operate on fixed‑sized ten... |
test-lite | jah-auto-t75/break-filter-js-from-html__XaV6rcT | jah-auto-t75 | t2 | break-filter-js-from-html | break-filter-js-from-html__XaV6rcT | 14 | 2026-10-15T00:18:01.380+00:00 | 0.95 | strong | hard | TASK: You are an AI assistant tasked with solving command-line tasks in a Linux environment. You will be given a task description and the output from previously executed commands. Your goal is to solve the task by providing batches of shell commands. Format your response as JSON with the following structure: { "analysi... |
test-lite | jah-auto-t75/gpt2-codegolf__wbNvMPD | jah-auto-t75 | pi | gpt2-codegolf | gpt2-codegolf__wbNvMPD | 26 | 2026-10-15T00:18:03.399+00:00 | 0.72 | fast | routine | TASK: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I will run it /app/a.out gp... |
test-lite | jah-auto-t75/gpt2-codegolf__wbNvMPD | jah-auto-t75 | pi | gpt2-codegolf | gpt2-codegolf__wbNvMPD | 27 | 2026-10-15T00:18:06.791+00:00 | 0.65 | fast | routine | TASK: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I will run it /app/a.out gp... |
test-lite | jah-auto-t75/break-filter-js-from-html__XaV6rcT | jah-auto-t75 | t2 | break-filter-js-from-html | break-filter-js-from-html__XaV6rcT | 15 | 2026-10-15T00:18:09.498+00:00 | 0.95 | strong | hard | TASK: You are an AI assistant tasked with solving command-line tasks in a Linux environment. You will be given a task description and the output from previously executed commands. Your goal is to solve the task by providing batches of shell commands. Format your response as JSON with the following structure: { "analysi... |
test-lite | jah-auto-t75/reshard-c4-data__yaojNSw | jah-auto-t75 | pi | reshard-c4-data | reshard-c4-data__yaojNSw | 28 | 2026-10-15T00:18:10.306+00:00 | 0.65 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/gpt2-codegolf__wbNvMPD | jah-auto-t75 | pi | gpt2-codegolf | gpt2-codegolf__wbNvMPD | 28 | 2026-10-15T00:18:12.356+00:00 | 0.8 | strong | hard | TASK: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I will run it /app/a.out gp... |
test-lite | jah-auto-t75/gpt2-codegolf__D5XWnZ6 | jah-auto-t75 | mswe | gpt2-codegolf | gpt2-codegolf__D5XWnZ6 | 20 | 2026-10-15T00:18:18.022+00:00 | 0.6 | fast | routine | TASK: Please solve this issue: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I ... |
test-lite | jah-auto-t75/write-compressor__MqbbAej | jah-auto-t75 | pi | write-compressor | write-compressor__MqbbAej | 11 | 2026-10-15T00:18:22.510+00:00 | 0.9 | strong | hard | TASK: I have a decompressor in /app/decomp.c. It reads compressed data from stdin and writes the decompressed data to stdout. I also have a file /app/data.txt that has a bunch of text. Write me data.comp that's compressed such that running cat data.comp | /app/decomp gives exactly data.txt. You can generate data.comp a... |
test-lite | jah-auto-t75/gpt2-codegolf__D5XWnZ6 | jah-auto-t75 | mswe | gpt2-codegolf | gpt2-codegolf__D5XWnZ6 | 21 | 2026-10-15T00:18:22.626+00:00 | 0.93 | strong | hard | TASK: Please solve this issue: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I ... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 27 | 2026-10-15T00:18:25.429+00:00 | 0.67 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/write-compressor__5cP4Man | jah-auto-t75 | mswe | write-compressor | write-compressor__5cP4Man | 8 | 2026-10-15T00:18:26.055+00:00 | 0.89 | strong | hard | TASK: Please solve this issue: I have a decompressor in /app/decomp.c. It reads compressed data from stdin and writes the decompressed data to stdout. I also have a file /app/data.txt that has a bunch of text. Write me data.comp that's compressed such that running cat data.comp | /app/decomp gives exactly data.txt. You... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 28 | 2026-10-15T00:18:26.704+00:00 | 0.59 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/reshard-c4-data__yaojNSw | jah-auto-t75 | pi | reshard-c4-data | reshard-c4-data__yaojNSw | 29 | 2026-10-15T00:18:27.312+00:00 | 0.46 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/write-compressor__5cP4Man | jah-auto-t75 | mswe | write-compressor | write-compressor__5cP4Man | 9 | 2026-10-15T00:18:47.963+00:00 | 0.87 | strong | hard | TASK: Please solve this issue: I have a decompressor in /app/decomp.c. It reads compressed data from stdin and writes the decompressed data to stdout. I also have a file /app/data.txt that has a bunch of text. Write me data.comp that's compressed such that running cat data.comp | /app/decomp gives exactly data.txt. You... |
test-lite | jah-auto-t75/merge-diff-arc-agi-task__gMQHLVC | jah-auto-t75 | pi | merge-diff-arc-agi-task | merge-diff-arc-agi-task__gMQHLVC | 1 | 2026-10-15T00:18:48.609+00:00 | 0.01 | fast | routine | TASK: mkdir /app/repo, then initialize a git repo at /app/repo. Fetch the first git bundle located at /app/bundle1.bundle and ensure it is checked out into a local branch named branch1, fetching from the HEAD reference, Fetch the second git bundle located at /app/bundle2.bundle and ensure it is checked out into a local... |
test-lite | jah-auto-t75/merge-diff-arc-agi-task__gMQHLVC | jah-auto-t75 | pi | merge-diff-arc-agi-task | merge-diff-arc-agi-task__gMQHLVC | 2 | 2026-10-15T00:18:49.357+00:00 | 0.98 | strong | hard | TASK: mkdir /app/repo, then initialize a git repo at /app/repo. Fetch the first git bundle located at /app/bundle1.bundle and ensure it is checked out into a local branch named branch1, fetching from the HEAD reference, Fetch the second git bundle located at /app/bundle2.bundle and ensure it is checked out into a local... |
test-lite | jah-auto-t75/gpt2-codegolf__ij4jD4w | jah-auto-t75 | dsh | gpt2-codegolf | gpt2-codegolf__ij4jD4w | 31 | 2026-10-15T00:18:51.217+00:00 | 0.76 | strong | hard | TASK: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I will run it /app/a.out gp... |
test-lite | jah-auto-t75/merge-diff-arc-agi-task__gMQHLVC | jah-auto-t75 | pi | merge-diff-arc-agi-task | merge-diff-arc-agi-task__gMQHLVC | 3 | 2026-10-15T00:18:56.436+00:00 | 0.92 | strong | hard | TASK: mkdir /app/repo, then initialize a git repo at /app/repo. Fetch the first git bundle located at /app/bundle1.bundle and ensure it is checked out into a local branch named branch1, fetching from the HEAD reference, Fetch the second git bundle located at /app/bundle2.bundle and ensure it is checked out into a local... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__vzKDdmM | jah-auto-t75 | dsh | llm-inference-batching-scheduler | llm-inference-batching-scheduler__vzKDdmM | 19 | 2026-10-15T00:19:01.345+00:00 | 0.81 | strong | hard | TASK: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to operate on fixed‑sized ten... |
test-lite | jah-auto-t75/gpt2-codegolf__wbNvMPD | jah-auto-t75 | pi | gpt2-codegolf | gpt2-codegolf__wbNvMPD | 29 | 2026-10-15T00:19:02.705+00:00 | 0.59 | fast | routine | TASK: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I will run it /app/a.out gp... |
test-lite | jah-auto-t75/gpt2-codegolf__wbNvMPD | jah-auto-t75 | pi | gpt2-codegolf | gpt2-codegolf__wbNvMPD | 30 | 2026-10-15T00:19:07.717+00:00 | 0.72 | fast | routine | TASK: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I will run it /app/a.out gp... |
test-lite | jah-auto-t75/merge-diff-arc-agi-task__gMQHLVC | jah-auto-t75 | pi | merge-diff-arc-agi-task | merge-diff-arc-agi-task__gMQHLVC | 4 | 2026-10-15T00:19:08.534+00:00 | 0.91 | strong | hard | TASK: mkdir /app/repo, then initialize a git repo at /app/repo. Fetch the first git bundle located at /app/bundle1.bundle and ensure it is checked out into a local branch named branch1, fetching from the HEAD reference, Fetch the second git bundle located at /app/bundle2.bundle and ensure it is checked out into a local... |
test-lite | jah-auto-t75/gpt2-codegolf__wbNvMPD | jah-auto-t75 | pi | gpt2-codegolf | gpt2-codegolf__wbNvMPD | 31 | 2026-10-15T00:19:08.812+00:00 | 0.89 | strong | hard | TASK: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I will run it /app/a.out gp... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 29 | 2026-10-15T00:19:10.676+00:00 | 0.53 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 30 | 2026-10-15T00:19:13.146+00:00 | 0.07 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 31 | 2026-10-15T00:19:14.879+00:00 | 0.06 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 32 | 2026-10-15T00:19:15.835+00:00 | 0.02 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/reshard-c4-data__yaojNSw | jah-auto-t75 | pi | reshard-c4-data | reshard-c4-data__yaojNSw | 30 | 2026-10-15T00:19:16.680+00:00 | 0.51 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__vzKDdmM | jah-auto-t75 | dsh | llm-inference-batching-scheduler | llm-inference-batching-scheduler__vzKDdmM | 20 | 2026-10-15T00:19:23.182+00:00 | 0.33 | fast | routine | TASK: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to operate on fixed‑sized ten... |
test-lite | jah-auto-t75/reshard-c4-data__ZPzsSd6 | jah-auto-t75 | dsh | reshard-c4-data | reshard-c4-data__ZPzsSd6 | 33 | 2026-10-15T00:19:23.336+00:00 | 0.01 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/llm-inference-batching-scheduler__vzKDdmM | jah-auto-t75 | dsh | llm-inference-batching-scheduler | llm-inference-batching-scheduler__vzKDdmM | 21 | 2026-10-15T00:19:27.267+00:00 | 0.77 | strong | hard | TASK: In this task you will implement an **LLM inference batching scheduler (shape‑aware)** for a static graph LLM inference system. **Background ** When running large language models on hardware accelerators (such as TPUs, RDUs, or bespoke inference chips) the compiled execution graph has to operate on fixed‑sized ten... |
test-lite | jah-auto-t75/reshard-c4-data__yaojNSw | jah-auto-t75 | pi | reshard-c4-data | reshard-c4-data__yaojNSw | 31 | 2026-10-15T00:19:28.590+00:00 | 0.15 | fast | routine | TASK: Help me create two scripts for managing the resharding of my dataset: 1. **/app/compress.py**: A script that takes an input directory and output directory as command-line arguments and reshards the data according to the following constraints: - Maximum 30 files or folders in each directory - Maximum 15MB filesize... |
test-lite | jah-auto-t75/gpt2-codegolf__ij4jD4w | jah-auto-t75 | dsh | gpt2-codegolf | gpt2-codegolf__ij4jD4w | 32 | 2026-10-15T00:19:31.387+00:00 | 0.89 | strong | hard | TASK: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I will run it /app/a.out gp... |
test-lite | jah-auto-t75/merge-diff-arc-agi-task__gMQHLVC | jah-auto-t75 | pi | merge-diff-arc-agi-task | merge-diff-arc-agi-task__gMQHLVC | 5 | 2026-10-15T00:19:37.008+00:00 | 0.1 | fast | routine | TASK: mkdir /app/repo, then initialize a git repo at /app/repo. Fetch the first git bundle located at /app/bundle1.bundle and ensure it is checked out into a local branch named branch1, fetching from the HEAD reference, Fetch the second git bundle located at /app/bundle2.bundle and ensure it is checked out into a local... |
test-lite | jah-auto-t75/merge-diff-arc-agi-task__gMQHLVC | jah-auto-t75 | pi | merge-diff-arc-agi-task | merge-diff-arc-agi-task__gMQHLVC | 6 | 2026-10-15T00:19:37.972+00:00 | 0.05 | fast | routine | TASK: mkdir /app/repo, then initialize a git repo at /app/repo. Fetch the first git bundle located at /app/bundle1.bundle and ensure it is checked out into a local branch named branch1, fetching from the HEAD reference, Fetch the second git bundle located at /app/bundle2.bundle and ensure it is checked out into a local... |
test-lite | jah-auto-t75/merge-diff-arc-agi-task__gMQHLVC | jah-auto-t75 | pi | merge-diff-arc-agi-task | merge-diff-arc-agi-task__gMQHLVC | 7 | 2026-10-15T00:19:39.227+00:00 | 0.06 | fast | routine | TASK: mkdir /app/repo, then initialize a git repo at /app/repo. Fetch the first git bundle located at /app/bundle1.bundle and ensure it is checked out into a local branch named branch1, fetching from the HEAD reference, Fetch the second git bundle located at /app/bundle2.bundle and ensure it is checked out into a local... |
test-lite | jah-auto-t75/merge-diff-arc-agi-task__gMQHLVC | jah-auto-t75 | pi | merge-diff-arc-agi-task | merge-diff-arc-agi-task__gMQHLVC | 8 | 2026-10-15T00:19:40.365+00:00 | 0.15 | fast | routine | TASK: mkdir /app/repo, then initialize a git repo at /app/repo. Fetch the first git bundle located at /app/bundle1.bundle and ensure it is checked out into a local branch named branch1, fetching from the HEAD reference, Fetch the second git bundle located at /app/bundle2.bundle and ensure it is checked out into a local... |
test-lite | jah-auto-t75/gpt2-codegolf__D5XWnZ6 | jah-auto-t75 | mswe | gpt2-codegolf | gpt2-codegolf__D5XWnZ6 | 22 | 2026-10-15T00:19:40.717+00:00 | 0.84 | strong | hard | TASK: Please solve this issue: I have downloaded the gpt-2 weights stored as a TF .ckpt. Write me a dependency-free C file that samples from the model with arg-max sampling. Call your program /app/gpt2.c, I will compile with gcc -O3 -lm. It should read the .ckpt and the .bpe file. Your c program must be <5000 bytes. I ... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.