SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2409.18073 · 2024

Infer Human's Intentions Before Following Natural Language Instructions

arXiv · PDF · Open in the Atlas

Code that ran

10 of 11 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 1 violated it and 0 have no contract verdict; 8 failures of our instrument; 1 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2409.18073.json · badge: /api/ran/2409.18073.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
simon-wan/fiser canonical (classifier label)
classifier basis: census
10 ran · 1 no recorded run
FunctionStatusWhere it lives
attention Ran
ran; our contract draft was wrong, not the code
simon-wan/fiser/networks/layers.py
code served (permissive licence) · get_code("ba5496494629bdbe")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
import math

torch.manual_seed(0)

# Create tiny tensors for attention computation
# batch_size=2, seq_len=3, d_k=4
batch_size = 2
seq_len = 3
d_k = 4

query = torch.randn(batch_size, seq_len, d_k)
key = torch.randn(batch_size, seq_len, d_k)
value = torch.randn(batch_size, seq_len, d_k)

# Create a mask (shape: batch_size, seq_len)
# 1 means keep, 0 means mask out
mask = torch.ones(batch_size, seq_len, dtype=torch.uint8)

# Create a dropout layer
dropout = nn.Dropout(p=0.1)

ARGS = {
    "query": query,
    "key": key,
    "value": value,
    "mask": mask,
    "dropout": dropout
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
clones Ran
ran; our contract draft was wrong, not the code
simon-wan/fiser/networks/layers.py
code served (permissive licence) · get_code("a3722169bbc81569")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned ModuleList. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
import copy

torch.manual_seed(0)

# Create a simple module to clone
class SimpleModule(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, x):
        return self.linear(x)

module = SimpleModule()
N = 5

ARGS = {
    "module": module,
    "N": N
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ModuleList — a ModuleList output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ModuleList output the recorded ndim 1 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_argument_index Ran
ran; our contract draft was wrong, not the code
simon-wan/fiser/dataloaders/data_utils.py
code served (permissive licence) · get_code("d0e451651bc94233")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

ARGS = {
    "arg": "  example_object  ",
    "object_names": ["example_object", "another_object", "third_object#0"]
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_embeddings Ran
ran; our contract draft was wrong, not the code
simon-wan/fiser/networks/embeddings.py
code served (permissive licence) · get_code("16b037202af1a18b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Create small vocabularies as lists
src_vocab = ['<pad>', 'the', 'cat', 'sat']
tgt_vocab = ['<pad>', 'dog', 'ran', 'jumped']

mid_vocabs = {
    'q': ['<pad>', 'what', 'where', 'who'],
    's': ['<pad>', 'subject', 'object', 'verb'],
    'v': ['<pad>', 'action1', 'action2', 'action3'],
    'o': ['<pad>', 'obj1', 'obj2', 'obj3']
}

d_model = 4
num_attr = 3
device = 'cpu'

ARGS = {
    'src_vocab': src_vocab,
    'tgt_vocab': tgt_vocab,
    'mid_vocabs': mid_vocabs,
    'd_model': d_model,
    'num_attr': num_attr,
    'device': device
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_loss_compute Ran
ran; our contract draft was wrong, not the code
simon-wan/fiser/networks/loss.py
code served (permissive licence) · get_code("1bcfd65891604bb3")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned LossCompute. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

class SimpleGenerator(nn.Module):
    """Minimal generator that outputs act_x, arg1_x, arg2_x"""
    def __init__(self):
        super().__init__()
        self.act_proj = nn.Linear(8, 4)
        self.arg1_proj = nn.Linear(8, 3)
        self.arg2_proj = nn.Linear(8, 5)
    
    def forward(self, x):
        act_x = self.act_proj(x)
        arg1_x = self.arg1_proj(x)
        arg2_x = self.arg2_proj(x)
        return act_x, arg1_x, arg2_x

generator = SimpleGenerator()

ARGS = {
    "generator": generator
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type LossCompute — a LossCompute output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a LossCompute output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_tokenizers_and_vocabs Ran
ran; our contract draft was wrong, not the code
simon-wan/fiser/networks/tokenizer.py
code served (permissive licence) · get_code("89dae777cd40b2a9")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import tempfile
import torch

# Create temporary directory with vocab files
temp_dir = tempfile.mkdtemp()

# Create minimal vocab files
vocab_files = {
    'src_vocab.txt': ['hello', 'world', 'test', 'code'],
    'tgt_vocab.txt': ['bonjour', 'monde', 'essai', 'code'],
    'mid_q_vocab.txt': ['query', 'question', 'ask'],
    'mid_s_vocab.txt': ['subject', 'topic', 'theme'],
    'mid_v_vocab.txt': ['verb', 'action', 'do'],
    'mid_o_vocab.txt': ['object', 'thing', 'item']
}

for filename, words in vocab_files.items():
    filepath = os.path.join(temp_dir, filename)
    with open(filepath, 'w') as f:
        f.write('\n'.join(words))

ARGS = {
    'vocabulary_dir': temp_dir
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
mid_loss_compute Ran
ran; our contract draft was wrong, not the code
simon-wan/fiser/networks/loss.py
code served (permissive licence) · get_code("688c7f03d5f4d95a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Create dummy outputs: tuple of (pred_q, pred_s, pred_v, pred_o)
# Each is a tensor of shape (batch_size, num_classes)
# Using batch_size=2 and num_classes=4
batch_size = 2
num_classes = 4

pred_q = torch.randn(batch_size, num_classes)
pred_s = torch.randn(batch_size, num_classes)
pred_v = torch.randn(batch_size, num_classes)
pred_o = torch.randn(batch_size, num_classes)

outputs = (pred_q, pred_s, pred_v, pred_o)

# Create dummy targets: tuple of (targ_q, targ_s, targ_v, targ_o)
# Each is a tensor of shape (batch_size,) with class indices
targ_q = torch.randint(0, num_classes, (batch_size,))
targ_s = torch.randint(0, num_classes, (batch_size,))
targ_v = torch.randint(0, num_classes, (batch_size,))
targ_o = torch.randint(0, num_classes, (batch_size,))

targets = (targ_q, targ_s, targ_v, targ_o)

ARGS = {
    "outputs": outputs,
    "targets": targets
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
obj_loss_compute Ran
ran; our contract draft was wrong, not the code
simon-wan/fiser/networks/loss.py
code served (permissive licence) · get_code("00cd233c89bd53a3")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# outputs: tensor of shape (batch_size, num_classes)
# The function calls criterion(outputs[i], target) which expects:
# - outputs[i]: 1D tensor of logits for a single sample
# - target: scalar or 1D tensor with class index
# Then it calls outputs.argmax(dim=-1) expecting 2D tensor

outputs = torch.randn(2, 4)  # batch_size=2, num_classes=4

# targets: must have shape[0] attribute (used in division)
# The function iterates through targets with enumerate
# Each target is passed to CrossEntropyLoss which expects a class index (long tensor)
targets = torch.tensor([1, 3])  # batch_size=2, class indices

ARGS = {
    "outputs": outputs,
    "targets": targets,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
parse_action Ran
ran, honoured the contract we drafted
simon-wan/fiser/dataloaders/data_utils.py
code served (permissive licence) · get_code("7a7525f5ae8ab255")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Based on the function body, parse_action takes:
# 1. action: a string describing an action
# 2. object_names: a list of object name strings

# The function parses various action types and extracts object names from the action string.
# It uses get_argument_index which searches for exact matches or appends '#0' to find matches.

# Create test data
action = "pick up apple from table"
object_names = ["apple#0", "table#0", "knife#0", "plate#0"]

ARGS = {
    "action": action,
    "object_names": object_names,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
subsequent_mask Ran
ran, violated the contract we drafted
simon-wan/fiser/networks/layers.py
code served (permissive licence) · get_code("50974789766576a9")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 3, dtype family float, shape (1, 5, 5) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

ARGS = {
    "size": 5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (1, 5, 5) — shape (1, 5, 5) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
parse_expert_demo Not yet run
unverified (harvested, no recorded run)
simon-wan/fiser/dataloaders/data_utils.py
code served (permissive licence) · get_code("d5e7f57c132f1578")
How it ran

no invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 1 the same month, 1 later) and 10 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
clones 79 other papers
66 dated earlier, 1 the same month, 11 later, 1 same year, order not determinable · 11 only as an assembly brick
1411.4555 (2014-11)
Show and Tell: A Neural Image Caption Generator
1612.00563 (2016-12)
Self-critical Sequence Training for Image Captioning
1706.03762 (2017-06)
Attention Is All You Need
and 76 more not named here

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

For AI agents to be helpful to humans, they should be able to follow natural language instructions to complete everyday cooperative tasks in human environments. However, real human instructions inherently possess ambiguity, because the human speakers assume sufficient prior knowledge about their hidden goals and intentions. Standard language grounding and planning methods fail to address such ambiguities because they do not model human internal goals as additional partially observable factors in the environment. We propose a new framework, Follow Instructions with Social and Embodied Reasoning (FISER), aiming for better natural language instruction following in collaborative embodied tasks. Our framework makes explicit inferences about human goals and intentions as intermediate reasoning steps. We implement a set of Transformer-based models and evaluate them over a challenging benchmark, HandMeThat. We empirically demonstrate that using social reasoning to explicitly infer human intentions before making action plans surpasses purely end-to-end approaches. We also compare our implementation with strong baselines, including Chain of Thought prompting on the largest available pre-trained language models, and find that FISER provides better performance on the embodied social reasoning tasks under investigation, reaching the state-of-the-art on HandMeThat.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2409.18073")
get_code_for_paper("2409.18073")
have("2409.18073")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2409.18073.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2409.18073.svg)](https://syntology.ai/paper/2409.18073)

Connect an agent — have() is free.