# research-code-bench

- domain: swe
- n_tasks: 212
- n_runnable: 0

## Tasks

| task | difficulty | language | runnable | harnesses tried |
|---|---|---|---|---|
| [advantage-alignment_aa_terms1](https://harnessreport.com/tasks/research-code-bench/advantage-alignment_aa_terms1.md) | difficult |  |  | 0 |
| [advantage-alignment_aa_terms2](https://harnessreport.com/tasks/research-code-bench/advantage-alignment_aa_terms2.md) | difficult |  |  | 0 |
| [advantage-alignment_aa_terms3](https://harnessreport.com/tasks/research-code-bench/advantage-alignment_aa_terms3.md) | difficult |  |  | 0 |
| [advantage-alignment_integrated_aa](https://harnessreport.com/tasks/research-code-bench/advantage-alignment_integrated_aa.md) | difficult |  |  | 0 |
| [advantage-alignment_proximal_surrogate](https://harnessreport.com/tasks/research-code-bench/advantage-alignment_proximal_surrogate.md) | difficult |  |  | 0 |
| [diff-transformer_empirical_lambda_init_function](https://harnessreport.com/tasks/research-code-bench/diff-transformer_empirical_lambda_init_function.md) | difficult |  |  | 0 |
| [diff-transformer_groupnorm_with_scaling](https://harnessreport.com/tasks/research-code-bench/diff-transformer_groupnorm_with_scaling.md) | difficult |  |  | 0 |
| [diff-transformer_lambda_setup](https://harnessreport.com/tasks/research-code-bench/diff-transformer_lambda_setup.md) | difficult |  |  | 0 |
| [diff-transformer_multihead_differential_attention](https://harnessreport.com/tasks/research-code-bench/diff-transformer_multihead_differential_attention.md) | difficult |  |  | 0 |
| [diff-transformer_normalization_setup](https://harnessreport.com/tasks/research-code-bench/diff-transformer_normalization_setup.md) | difficult |  |  | 0 |
| [diff-transformer_re-parameterize_lambda](https://harnessreport.com/tasks/research-code-bench/diff-transformer_re-parameterize_lambda.md) | difficult |  |  | 0 |
| [diff-transformer_taking_difference](https://harnessreport.com/tasks/research-code-bench/diff-transformer_taking_difference.md) | difficult |  |  | 0 |
| [diffusiondpo_calculate_model_losses](https://harnessreport.com/tasks/research-code-bench/diffusiondpo_calculate_model_losses.md) | difficult |  |  | 0 |
| [diffusiondpo_calculate_reference_losses](https://harnessreport.com/tasks/research-code-bench/diffusiondpo_calculate_reference_losses.md) | difficult |  |  | 0 |
| [diffusiondpo_dpo_loss_computation](https://harnessreport.com/tasks/research-code-bench/diffusiondpo_dpo_loss_computation.md) | difficult |  |  | 0 |
| [diffusiondpo_dpo_loss_formula](https://harnessreport.com/tasks/research-code-bench/diffusiondpo_dpo_loss_formula.md) | difficult |  |  | 0 |
| [diffusiondpo_implicit_accuracy_metric](https://harnessreport.com/tasks/research-code-bench/diffusiondpo_implicit_accuracy_metric.md) | difficult |  |  | 0 |
| [diffusiondpo_loss_computation](https://harnessreport.com/tasks/research-code-bench/diffusiondpo_loss_computation.md) | difficult |  |  | 0 |
| [diffusiondpo_raw_model_loss_metric](https://harnessreport.com/tasks/research-code-bench/diffusiondpo_raw_model_loss_metric.md) | difficult |  |  | 0 |
| [diffusiondpo_raw_ref_loss](https://harnessreport.com/tasks/research-code-bench/diffusiondpo_raw_ref_loss.md) | difficult |  |  | 0 |
| [diffusiondpo_sft_loss](https://harnessreport.com/tasks/research-code-bench/diffusiondpo_sft_loss.md) | difficult |  |  | 0 |
| [dyt_alpha](https://harnessreport.com/tasks/research-code-bench/dyt_alpha.md) | difficult |  |  | 0 |
| [dyt_bias](https://harnessreport.com/tasks/research-code-bench/dyt_bias.md) | difficult |  |  | 0 |
| [dyt_channel_split](https://harnessreport.com/tasks/research-code-bench/dyt_channel_split.md) | difficult |  |  | 0 |
| [dyt_channels_first](https://harnessreport.com/tasks/research-code-bench/dyt_channels_first.md) | difficult |  |  | 0 |
| [dyt_channels_last](https://harnessreport.com/tasks/research-code-bench/dyt_channels_last.md) | difficult |  |  | 0 |
| [dyt_forward](https://harnessreport.com/tasks/research-code-bench/dyt_forward.md) | difficult |  |  | 0 |
| [dyt_init](https://harnessreport.com/tasks/research-code-bench/dyt_init.md) | difficult |  |  | 0 |
| [dyt_tanh](https://harnessreport.com/tasks/research-code-bench/dyt_tanh.md) | difficult |  |  | 0 |
| [dyt_weight](https://harnessreport.com/tasks/research-code-bench/dyt_weight.md) | difficult |  |  | 0 |
| [eomt_activation](https://harnessreport.com/tasks/research-code-bench/eomt_activation.md) | difficult |  |  | 0 |
| [eomt_add_query_tokens](https://harnessreport.com/tasks/research-code-bench/eomt_add_query_tokens.md) | difficult |  |  | 0 |
| [eomt_apply_masking_probability](https://harnessreport.com/tasks/research-code-bench/eomt_apply_masking_probability.md) | difficult |  |  | 0 |
| [eomt_calculate_upscale_parameters](https://harnessreport.com/tasks/research-code-bench/eomt_calculate_upscale_parameters.md) | difficult |  |  | 0 |
| [eomt_convolution_1](https://harnessreport.com/tasks/research-code-bench/eomt_convolution_1.md) | difficult |  |  | 0 |
| [eomt_convolution_2](https://harnessreport.com/tasks/research-code-bench/eomt_convolution_2.md) | difficult |  |  | 0 |
| [eomt_create_attention_mask](https://harnessreport.com/tasks/research-code-bench/eomt_create_attention_mask.md) | difficult |  |  | 0 |
| [eomt_custom_attention_mechanism](https://harnessreport.com/tasks/research-code-bench/eomt_custom_attention_mechanism.md) | difficult |  |  | 0 |
| [eomt_disable_attention_mask](https://harnessreport.com/tasks/research-code-bench/eomt_disable_attention_mask.md) | difficult |  |  | 0 |
| [eomt_eomt_initialization](https://harnessreport.com/tasks/research-code-bench/eomt_eomt_initialization.md) | difficult |  |  | 0 |
| [eomt_extract_query_key_value](https://harnessreport.com/tasks/research-code-bench/eomt_extract_query_key_value.md) | difficult |  |  | 0 |
| [eomt_extract_query_tokens](https://harnessreport.com/tasks/research-code-bench/eomt_extract_query_tokens.md) | difficult |  |  | 0 |
| [eomt_final_prediction](https://harnessreport.com/tasks/research-code-bench/eomt_final_prediction.md) | difficult |  |  | 0 |
| [eomt_forward_pass](https://harnessreport.com/tasks/research-code-bench/eomt_forward_pass.md) | difficult |  |  | 0 |
| [eomt_generate_class_logits](https://harnessreport.com/tasks/research-code-bench/eomt_generate_class_logits.md) | difficult |  |  | 0 |
| [eomt_generate_mask_logits](https://harnessreport.com/tasks/research-code-bench/eomt_generate_mask_logits.md) | difficult |  |  | 0 |
| [eomt_generate_masks](https://harnessreport.com/tasks/research-code-bench/eomt_generate_masks.md) | difficult |  |  | 0 |
| [eomt_initialize_class_prediction_head](https://harnessreport.com/tasks/research-code-bench/eomt_initialize_class_prediction_head.md) | difficult |  |  | 0 |
| [eomt_initialize_mask_prediction_head](https://harnessreport.com/tasks/research-code-bench/eomt_initialize_mask_prediction_head.md) | difficult |  |  | 0 |
| [eomt_initialize_query_embedding](https://harnessreport.com/tasks/research-code-bench/eomt_initialize_query_embedding.md) | difficult |  |  | 0 |
| [eomt_initialize_tracking_variables](https://harnessreport.com/tasks/research-code-bench/eomt_initialize_tracking_variables.md) | difficult |  |  | 0 |
| [eomt_initialize_upscale_module](https://harnessreport.com/tasks/research-code-bench/eomt_initialize_upscale_module.md) | difficult |  |  | 0 |
| [eomt_input_normalization](https://harnessreport.com/tasks/research-code-bench/eomt_input_normalization.md) | difficult |  |  | 0 |
| [eomt_normalization](https://harnessreport.com/tasks/research-code-bench/eomt_normalization.md) | difficult |  |  | 0 |
| [eomt_patch_embedding](https://harnessreport.com/tasks/research-code-bench/eomt_patch_embedding.md) | difficult |  |  | 0 |
| [eomt_predict_from_features](https://harnessreport.com/tasks/research-code-bench/eomt_predict_from_features.md) | difficult |  |  | 0 |
| [eomt_predict_masks_and_classes](https://harnessreport.com/tasks/research-code-bench/eomt_predict_masks_and_classes.md) | difficult |  |  | 0 |
| [eomt_process_attention_mask](https://harnessreport.com/tasks/research-code-bench/eomt_process_attention_mask.md) | difficult |  |  | 0 |
| [eomt_process_image_features](https://harnessreport.com/tasks/research-code-bench/eomt_process_image_features.md) | difficult |  |  | 0 |
| [eomt_process_transformer_blocks](https://harnessreport.com/tasks/research-code-bench/eomt_process_transformer_blocks.md) | difficult |  |  | 0 |
| [eomt_project_attention_output](https://harnessreport.com/tasks/research-code-bench/eomt_project_attention_output.md) | difficult |  |  | 0 |
| [eomt_random_mask_disable](https://harnessreport.com/tasks/research-code-bench/eomt_random_mask_disable.md) | difficult |  |  | 0 |
| [eomt_register_attention_mask_probabilities](https://harnessreport.com/tasks/research-code-bench/eomt_register_attention_mask_probabilities.md) | difficult |  |  | 0 |
| [eomt_scale_block_forward](https://harnessreport.com/tasks/research-code-bench/eomt_scale_block_forward.md) | difficult |  |  | 0 |
| [eomt_scale_block_initialization](https://harnessreport.com/tasks/research-code-bench/eomt_scale_block_initialization.md) | difficult |  |  | 0 |
| [eomt_store_parameters](https://harnessreport.com/tasks/research-code-bench/eomt_store_parameters.md) | difficult |  |  | 0 |
| [eomt_transformer_block_forward](https://harnessreport.com/tasks/research-code-bench/eomt_transformer_block_forward.md) | difficult |  |  | 0 |
| [fractalgen_cfg_schedule](https://harnessreport.com/tasks/research-code-bench/fractalgen_cfg_schedule.md) | difficult |  |  | 0 |
| [fractalgen_chunk_mask_to_pred](https://harnessreport.com/tasks/research-code-bench/fractalgen_chunk_mask_to_pred.md) | difficult |  |  | 0 |
| [fractalgen_concatenate_patches](https://harnessreport.com/tasks/research-code-bench/fractalgen_concatenate_patches.md) | difficult |  |  | 0 |
| [fractalgen_init_token_mask__patches__orders__num_iter](https://harnessreport.com/tasks/research-code-bench/fractalgen_init_token_mask__patches__orders__num_iter.md) | difficult |  |  | 0 |
| [fractalgen_mask_by_order](https://harnessreport.com/tasks/research-code-bench/fractalgen_mask_by_order.md) | difficult |  |  | 0 |
| [fractalgen_mask_ratio_and_mask_length](https://harnessreport.com/tasks/research-code-bench/fractalgen_mask_ratio_and_mask_length.md) | difficult |  |  | 0 |
| [fractalgen_next_level_sample_function](https://harnessreport.com/tasks/research-code-bench/fractalgen_next_level_sample_function.md) | difficult |  |  | 0 |
| [fractalgen_patchify](https://harnessreport.com/tasks/research-code-bench/fractalgen_patchify.md) | difficult |  |  | 0 |
| [fractalgen_random_masking](https://harnessreport.com/tasks/research-code-bench/fractalgen_random_masking.md) | difficult |  |  | 0 |
| [fractalgen_random_masking_uniform](https://harnessreport.com/tasks/research-code-bench/fractalgen_random_masking_uniform.md) | difficult |  |  | 0 |
| [fractalgen_sample_image](https://harnessreport.com/tasks/research-code-bench/fractalgen_sample_image.md) | difficult |  |  | 0 |
| [fractalgen_sample_orders](https://harnessreport.com/tasks/research-code-bench/fractalgen_sample_orders.md) | difficult |  |  | 0 |
| [fractalgen_unpatchify](https://harnessreport.com/tasks/research-code-bench/fractalgen_unpatchify.md) | difficult |  |  | 0 |
| [gmflow_apply_2nd_order_gm](https://harnessreport.com/tasks/research-code-bench/gmflow_apply_2nd_order_gm.md) | difficult |  |  | 0 |
| [gmflow_apply_mask_to_gm](https://harnessreport.com/tasks/research-code-bench/gmflow_apply_mask_to_gm.md) | difficult |  |  | 0 |
| [gmflow_apply_probabilistic_cfg](https://harnessreport.com/tasks/research-code-bench/gmflow_apply_probabilistic_cfg.md) | difficult |  |  | 0 |
| [gmflow_compute_final_nll](https://harnessreport.com/tasks/research-code-bench/gmflow_compute_final_nll.md) | difficult |  |  | 0 |
| [gmflow_compute_gaussian_log_likelihood](https://harnessreport.com/tasks/research-code-bench/gmflow_compute_gaussian_log_likelihood.md) | difficult |  |  | 0 |
| [gmflow_compute_gaussian_mask](https://harnessreport.com/tasks/research-code-bench/gmflow_compute_gaussian_mask.md) | difficult |  |  | 0 |
| [gmflow_compute_inverse_stds](https://harnessreport.com/tasks/research-code-bench/gmflow_compute_inverse_stds.md) | difficult |  |  | 0 |
| [gmflow_compute_noise_scale](https://harnessreport.com/tasks/research-code-bench/gmflow_compute_noise_scale.md) | difficult |  |  | 0 |
| [gmflow_compute_timestep_coefficients](https://harnessreport.com/tasks/research-code-bench/gmflow_compute_timestep_coefficients.md) | difficult |  |  | 0 |
| [gmflow_compute_weighted_difference](https://harnessreport.com/tasks/research-code-bench/gmflow_compute_weighted_difference.md) | difficult |  |  | 0 |
| [gmflow_convert_model_output_to_x0](https://harnessreport.com/tasks/research-code-bench/gmflow_convert_model_output_to_x0.md) | difficult |  |  | 0 |
| [gmflow_gm_kl_loss](https://harnessreport.com/tasks/research-code-bench/gmflow_gm_kl_loss.md) | difficult |  |  | 0 |
| [gmflow_main_sampling_loop](https://harnessreport.com/tasks/research-code-bench/gmflow_main_sampling_loop.md) | difficult |  |  | 0 |
| [gmflow_perform_gm_sdeode_step](https://harnessreport.com/tasks/research-code-bench/gmflow_perform_gm_sdeode_step.md) | difficult |  |  | 0 |
| [gmflow_prepare_timestep_tensor](https://harnessreport.com/tasks/research-code-bench/gmflow_prepare_timestep_tensor.md) | difficult |  |  | 0 |
| [gmflow_set_timesteps](https://harnessreport.com/tasks/research-code-bench/gmflow_set_timesteps.md) | difficult |  |  | 0 |
| [gps_chamfer_distance](https://harnessreport.com/tasks/research-code-bench/gps_chamfer_distance.md) | difficult |  |  | 0 |
| [gps_distance_between_two_sets](https://harnessreport.com/tasks/research-code-bench/gps_distance_between_two_sets.md) | difficult |  |  | 0 |
| [gps_euclidean_distance_computation](https://harnessreport.com/tasks/research-code-bench/gps_euclidean_distance_computation.md) | difficult |  |  | 0 |
| [gps_gps_reparameterization_with_minimum_euclidean_distance](https://harnessreport.com/tasks/research-code-bench/gps_gps_reparameterization_with_minimum_euclidean_distance.md) | difficult |  |  | 0 |
| [gps_gumbel_distribution_similarity_transformation](https://harnessreport.com/tasks/research-code-bench/gps_gumbel_distribution_similarity_transformation.md) | difficult |  |  | 0 |
| [gps_pairwise_expansion](https://harnessreport.com/tasks/research-code-bench/gps_pairwise_expansion.md) | difficult |  |  | 0 |
| [grid-cell-conformal-isometry__dx_to_theta_id_dr](https://harnessreport.com/tasks/research-code-bench/grid-cell-conformal-isometry__dx_to_theta_id_dr.md) | difficult |  |  | 0 |
| [grid-cell-conformal-isometry_activity_metrics](https://harnessreport.com/tasks/research-code-bench/grid-cell-conformal-isometry_activity_metrics.md) | difficult |  |  | 0 |
| [grid-cell-conformal-isometry_apply_additive_transformation](https://harnessreport.com/tasks/research-code-bench/grid-cell-conformal-isometry_apply_additive_transformation.md) | difficult |  |  | 0 |
| [grid-cell-conformal-isometry_apply_nonlinear_transform](https://harnessreport.com/tasks/research-code-bench/grid-cell-conformal-isometry_apply_nonlinear_transform.md) | difficult |  |  | 0 |
| [grid-cell-conformal-isometry_apply_transformation](https://harnessreport.com/tasks/research-code-bench/grid-cell-conformal-isometry_apply_transformation.md) | difficult |  |  | 0 |
| [grid-cell-conformal-isometry_calculate_distance_preservation](https://harnessreport.com/tasks/research-code-bench/grid-cell-conformal-isometry_calculate_distance_preservation.md) | difficult |  |  | 0 |
| [grid-cell-conformal-isometry_compute_neural_representations](https://harnessreport.com/tasks/research-code-bench/grid-cell-conformal-isometry_compute_neural_representations.md) | difficult |  |  | 0 |
| [grid-cell-conformal-isometry_compute_transformation_error](https://harnessreport.com/tasks/research-code-bench/grid-cell-conformal-isometry_compute_transformation_error.md) | difficult |  |  | 0 |
| [grid-cell-conformal-isometry_compute_transformation_matrix](https://harnessreport.com/tasks/research-code-bench/grid-cell-conformal-isometry_compute_transformation_matrix.md) | difficult |  |  | 0 |
| [grid-cell-conformal-isometry_conformal_isometry_loss](https://harnessreport.com/tasks/research-code-bench/grid-cell-conformal-isometry_conformal_isometry_loss.md) | difficult |  |  | 0 |
| [grid-cell-conformal-isometry_linear_transformation](https://harnessreport.com/tasks/research-code-bench/grid-cell-conformal-isometry_linear_transformation.md) | difficult |  |  | 0 |
| [grid-cell-conformal-isometry_nonlinear_transformation_1](https://harnessreport.com/tasks/research-code-bench/grid-cell-conformal-isometry_nonlinear_transformation_1.md) | difficult |  |  | 0 |
| [grid-cell-conformal-isometry_prepare_displacement](https://harnessreport.com/tasks/research-code-bench/grid-cell-conformal-isometry_prepare_displacement.md) | difficult |  |  | 0 |
| [grid-cell-conformal-isometry_prepare_inputs](https://harnessreport.com/tasks/research-code-bench/grid-cell-conformal-isometry_prepare_inputs.md) | difficult |  |  | 0 |
| [grid-cell-conformal-isometry_transformation_loss](https://harnessreport.com/tasks/research-code-bench/grid-cell-conformal-isometry_transformation_loss.md) | difficult |  |  | 0 |
| [hyla_mlp_linear](https://harnessreport.com/tasks/research-code-bench/hyla_mlp_linear.md) | difficult |  |  | 0 |
| [hyla_mlp_relu](https://harnessreport.com/tasks/research-code-bench/hyla_mlp_relu.md) | difficult |  |  | 0 |
| [hyla_rms_head](https://harnessreport.com/tasks/research-code-bench/hyla_rms_head.md) | difficult |  |  | 0 |
| [len_combine_gradients](https://harnessreport.com/tasks/research-code-bench/len_combine_gradients.md) | difficult |  |  | 0 |
| [len_compute_gradients](https://harnessreport.com/tasks/research-code-bench/len_compute_gradients.md) | difficult |  |  | 0 |
| [len_compute_schur_decomposition](https://harnessreport.com/tasks/research-code-bench/len_compute_schur_decomposition.md) | difficult |  |  | 0 |
| [len_construct_jacobian_matrix](https://harnessreport.com/tasks/research-code-bench/len_construct_jacobian_matrix.md) | difficult |  |  | 0 |
| [len_gamma_update_function](https://harnessreport.com/tasks/research-code-bench/len_gamma_update_function.md) | difficult |  |  | 0 |
| [len_gradient_descent_ascent_field](https://harnessreport.com/tasks/research-code-bench/len_gradient_descent_ascent_field.md) | difficult |  |  | 0 |
| [len_initialize_variables](https://harnessreport.com/tasks/research-code-bench/len_initialize_variables.md) | difficult |  |  | 0 |
| [len_jacobian_computation](https://harnessreport.com/tasks/research-code-bench/len_jacobian_computation.md) | difficult |  |  | 0 |
| [len_lazy_extra_newton](https://harnessreport.com/tasks/research-code-bench/len_lazy_extra_newton.md) | difficult |  |  | 0 |
| [len_main_iteration_loop](https://harnessreport.com/tasks/research-code-bench/len_main_iteration_loop.md) | difficult |  |  | 0 |
| [len_split_input_and_compute_norm](https://harnessreport.com/tasks/research-code-bench/len_split_input_and_compute_norm.md) | difficult |  |  | 0 |
| [len_split_input_vector](https://harnessreport.com/tasks/research-code-bench/len_split_input_vector.md) | difficult |  |  | 0 |
| [len_track_metrics](https://harnessreport.com/tasks/research-code-bench/len_track_metrics.md) | difficult |  |  | 0 |
| [len_update_steps](https://harnessreport.com/tasks/research-code-bench/len_update_steps.md) | difficult |  |  | 0 |
| [llm-sci-use_ai_text_probabilities](https://harnessreport.com/tasks/research-code-bench/llm-sci-use_ai_text_probabilities.md) | difficult |  |  | 0 |
| [llm-sci-use_alpha_estimation](https://harnessreport.com/tasks/research-code-bench/llm-sci-use_alpha_estimation.md) | difficult |  |  | 0 |
| [llm-sci-use_alpha_optimization](https://harnessreport.com/tasks/research-code-bench/llm-sci-use_alpha_optimization.md) | difficult |  |  | 0 |
| [llm-sci-use_bootstrap_inference](https://harnessreport.com/tasks/research-code-bench/llm-sci-use_bootstrap_inference.md) | difficult |  |  | 0 |
| [llm-sci-use_bootstrap_sampling](https://harnessreport.com/tasks/research-code-bench/llm-sci-use_bootstrap_sampling.md) | difficult |  |  | 0 |
| [llm-sci-use_confidence_interval_calculation](https://harnessreport.com/tasks/research-code-bench/llm-sci-use_confidence_interval_calculation.md) | difficult |  |  | 0 |
| [llm-sci-use_data_loading](https://harnessreport.com/tasks/research-code-bench/llm-sci-use_data_loading.md) | difficult |  |  | 0 |
| [llm-sci-use_data_preprocessing](https://harnessreport.com/tasks/research-code-bench/llm-sci-use_data_preprocessing.md) | difficult |  |  | 0 |
| [llm-sci-use_human_text_probabilities](https://harnessreport.com/tasks/research-code-bench/llm-sci-use_human_text_probabilities.md) | difficult |  |  | 0 |
| [llm-sci-use_inference_pipeline](https://harnessreport.com/tasks/research-code-bench/llm-sci-use_inference_pipeline.md) | difficult |  |  | 0 |
| [llm-sci-use_loading_token_probabilities](https://harnessreport.com/tasks/research-code-bench/llm-sci-use_loading_token_probabilities.md) | difficult |  |  | 0 |
| [llm-sci-use_log_likelihood_optimization](https://harnessreport.com/tasks/research-code-bench/llm-sci-use_log_likelihood_optimization.md) | difficult |  |  | 0 |
| [llm-sci-use_mixture_log_likelihood_calculation](https://harnessreport.com/tasks/research-code-bench/llm-sci-use_mixture_log_likelihood_calculation.md) | difficult |  |  | 0 |
| [llm-sci-use_precompute_probabilities](https://harnessreport.com/tasks/research-code-bench/llm-sci-use_precompute_probabilities.md) | difficult |  |  | 0 |
| [llm-sci-use_token_distribution_computation](https://harnessreport.com/tasks/research-code-bench/llm-sci-use_token_distribution_computation.md) | difficult |  |  | 0 |
| [minp_apply_filter_to_scores](https://harnessreport.com/tasks/research-code-bench/minp_apply_filter_to_scores.md) | difficult |  |  | 0 |
| [minp_convert_logits_to_probabilities](https://harnessreport.com/tasks/research-code-bench/minp_convert_logits_to_probabilities.md) | difficult |  |  | 0 |
| [minp_ensure_minimum_tokens_are_kept](https://harnessreport.com/tasks/research-code-bench/minp_ensure_minimum_tokens_are_kept.md) | difficult |  |  | 0 |
| [minp_find_maximum_probability_token](https://harnessreport.com/tasks/research-code-bench/minp_find_maximum_probability_token.md) | difficult |  |  | 0 |
| [minp_identify_tokens_to_remove](https://harnessreport.com/tasks/research-code-bench/minp_identify_tokens_to_remove.md) | difficult |  |  | 0 |
| [minp_min-p_sampling](https://harnessreport.com/tasks/research-code-bench/minp_min-p_sampling.md) | difficult |  |  | 0 |
| [minp_scale_min_p_threshold](https://harnessreport.com/tasks/research-code-bench/minp_scale_min_p_threshold.md) | difficult |  |  | 0 |
| [optimalsteps_dynamic_programming_1](https://harnessreport.com/tasks/research-code-bench/optimalsteps_dynamic_programming_1.md) | difficult |  |  | 0 |
| [optimalsteps_dynamic_programming_2](https://harnessreport.com/tasks/research-code-bench/optimalsteps_dynamic_programming_2.md) | difficult |  |  | 0 |
| [optimalsteps_dynamic_programming_3](https://harnessreport.com/tasks/research-code-bench/optimalsteps_dynamic_programming_3.md) | difficult |  |  | 0 |
| [optimalsteps_dynamic_programming_4](https://harnessreport.com/tasks/research-code-bench/optimalsteps_dynamic_programming_4.md) | difficult |  |  | 0 |
| [optimalsteps_dynamic_programming_for_each_sample](https://harnessreport.com/tasks/research-code-bench/optimalsteps_dynamic_programming_for_each_sample.md) | difficult |  |  | 0 |
| [optimalsteps_dynamic_programming_initialization_for_each_sample](https://harnessreport.com/tasks/research-code-bench/optimalsteps_dynamic_programming_initialization_for_each_sample.md) | difficult |  |  | 0 |
| [optimalsteps_dynamic_programming_search](https://harnessreport.com/tasks/research-code-bench/optimalsteps_dynamic_programming_search.md) | difficult |  |  | 0 |
| [optimalsteps_dynamic_programming_traceback](https://harnessreport.com/tasks/research-code-bench/optimalsteps_dynamic_programming_traceback.md) | difficult |  |  | 0 |
| [optimalsteps_search_oss_function](https://harnessreport.com/tasks/research-code-bench/optimalsteps_search_oss_function.md) | difficult |  |  | 0 |
| [optimalsteps_teacher_trajectory_computation](https://harnessreport.com/tasks/research-code-bench/optimalsteps_teacher_trajectory_computation.md) | difficult |  |  | 0 |
| [optimalsteps_teacher_trajectory_computation_for_each_timestep](https://harnessreport.com/tasks/research-code-bench/optimalsteps_teacher_trajectory_computation_for_each_timestep.md) | difficult |  |  | 0 |
| [repa-e_check1](https://harnessreport.com/tasks/research-code-bench/repa-e_check1.md) | difficult |  |  | 0 |
| [repa-e_check2](https://harnessreport.com/tasks/research-code-bench/repa-e_check2.md) | difficult |  |  | 0 |
| [repa-e_sit_loss](https://harnessreport.com/tasks/research-code-bench/repa-e_sit_loss.md) | difficult |  |  | 0 |
| [repa-e_sitbatchnorm_eps_is_set_to_1e-4](https://harnessreport.com/tasks/research-code-bench/repa-e_sitbatchnorm_eps_is_set_to_1e-4.md) | difficult |  |  | 0 |
| [repa-e_vae_loss](https://harnessreport.com/tasks/research-code-bench/repa-e_vae_loss.md) | difficult |  |  | 0 |
| [schedule_free_averaging_update](https://harnessreport.com/tasks/research-code-bench/schedule_free_averaging_update.md) | difficult |  |  | 0 |
| [schedule_free_averaging_update_and_interpolation](https://harnessreport.com/tasks/research-code-bench/schedule_free_averaging_update_and_interpolation.md) | difficult |  |  | 0 |
| [schedule_free_bias_correction](https://harnessreport.com/tasks/research-code-bench/schedule_free_bias_correction.md) | difficult |  |  | 0 |
| [schedule_free_decay_the_first_and_second_moment_running_average_coefficient](https://harnessreport.com/tasks/research-code-bench/schedule_free_decay_the_first_and_second_moment_running_average_coefficient.md) | difficult |  |  | 0 |
| [schedule_free_denom](https://harnessreport.com/tasks/research-code-bench/schedule_free_denom.md) | difficult |  |  | 0 |
| [schedule_free_dynamic_weighting](https://harnessreport.com/tasks/research-code-bench/schedule_free_dynamic_weighting.md) | difficult |  |  | 0 |
| [schedule_free_dynamic_weighting_and_averaging](https://harnessreport.com/tasks/research-code-bench/schedule_free_dynamic_weighting_and_averaging.md) | difficult |  |  | 0 |
| [schedule_free_gradient_evaluation_interpolation](https://harnessreport.com/tasks/research-code-bench/schedule_free_gradient_evaluation_interpolation.md) | difficult |  |  | 0 |
| [schedule_free_learning_rate](https://harnessreport.com/tasks/research-code-bench/schedule_free_learning_rate.md) | difficult |  |  | 0 |
| [schedule_free_moment_running_average_coefficient](https://harnessreport.com/tasks/research-code-bench/schedule_free_moment_running_average_coefficient.md) | difficult |  |  | 0 |
| [schedule_free_step](https://harnessreport.com/tasks/research-code-bench/schedule_free_step.md) | difficult |  |  | 0 |
| [schedule_free_warmup_schedule](https://harnessreport.com/tasks/research-code-bench/schedule_free_warmup_schedule.md) | difficult |  |  | 0 |
| [schedule_free_weight_decay](https://harnessreport.com/tasks/research-code-bench/schedule_free_weight_decay.md) | difficult |  |  | 0 |
| [schedule_free_z_update](https://harnessreport.com/tasks/research-code-bench/schedule_free_z_update.md) | difficult |  |  | 0 |
| [semanticist_apply_embedding_masks](https://harnessreport.com/tasks/research-code-bench/semanticist_apply_embedding_masks.md) | difficult |  |  | 0 |
| [semanticist_decode_multiple_tokens](https://harnessreport.com/tasks/research-code-bench/semanticist_decode_multiple_tokens.md) | difficult |  |  | 0 |
| [semanticist_decode_single_token](https://harnessreport.com/tasks/research-code-bench/semanticist_decode_single_token.md) | difficult |  |  | 0 |
| [semanticist_generate_main_function](https://harnessreport.com/tasks/research-code-bench/semanticist_generate_main_function.md) | difficult |  |  | 0 |
| [semanticist_generate_remaining_tokens](https://harnessreport.com/tasks/research-code-bench/semanticist_generate_remaining_tokens.md) | difficult |  |  | 0 |
| [semanticist_get_cfg_scale](https://harnessreport.com/tasks/research-code-bench/semanticist_get_cfg_scale.md) | difficult |  |  | 0 |
| [semanticist_initialize_sequence](https://harnessreport.com/tasks/research-code-bench/semanticist_initialize_sequence.md) | difficult |  |  | 0 |
| [semanticist_prefill_initial_tokens](https://harnessreport.com/tasks/research-code-bench/semanticist_prefill_initial_tokens.md) | difficult |  |  | 0 |
| [semanticist_prefill_sequence](https://harnessreport.com/tasks/research-code-bench/semanticist_prefill_sequence.md) | difficult |  |  | 0 |
| [semanticist_setup_conditions](https://harnessreport.com/tasks/research-code-bench/semanticist_setup_conditions.md) | difficult |  |  | 0 |
| [semanticist_setup_sequence_parameters](https://harnessreport.com/tasks/research-code-bench/semanticist_setup_sequence_parameters.md) | difficult |  |  | 0 |
| [siss_compute_loss_terms](https://harnessreport.com/tasks/research-code-bench/siss_compute_loss_terms.md) | difficult |  |  | 0 |
| [siss_create_defensive_mixture](https://harnessreport.com/tasks/research-code-bench/siss_create_defensive_mixture.md) | difficult |  |  | 0 |
| [siss_double_forward_passes](https://harnessreport.com/tasks/research-code-bench/siss_double_forward_passes.md) | difficult |  |  | 0 |

More: https://harnessreport.com/tasks/research-code-bench.md?after=siss_double_forward_passes
---
Harness Report runs agent harnesses from their GitHub repos on Harbor tasks and records every model call. Every page is also `.md` and `.json`; index: https://harnessreport.com/llms.txt · MCP: https://harnessreport.com/mcp